JPH07160074A - Electrophotographic process control device - Google Patents

Abstract

(57) Abstract: In short development period than Objective, to enable the construction of a control system for performing proper control in consideration of characteristics of the sensitive light body. In an electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, a charging potential, an exposed portion potential, a temperature, a humidity, a total copy number, Sensor means 28 for acquiring information affecting the electrophotographic process such as body rotation time and photoconductor deterioration degree, output data of this sensor means, charge amount such as charging charger voltage, charging grid voltage, and exposure. Parameter of the photoconductor obtained in advance based on experiments etc. by inputting at least one of parameters affecting the surface potential of the photoconductor such as exposure voltage such as driving voltage of the light source, pulse modulation width, etc. And a surface potential of the photoconductor as a teacher value, a latent image potential characteristic of the photoconductor is formed by a learned neural network to form a photoconductor surface. The photosensitive member surface potential estimating means 26 for estimating the surface potential is provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image forming apparatus such as a copying machine, a printer or a facsimile machine for copying or printing an image by an electrophotographic process mechanism including a process for forming an electrostatic latent image on a photoconductor. On the other hand, the present invention relates to an electrophotographic process control device for controlling each part thereof.

[0002]

2. Description of the Related Art Conventionally, as a latent image control method in this type of electrophotographic process, there are a table reference method and a feedback control method. The table reference method measures the surface potential on the photoconductor with a surface potentiometer, etc., and uses the table created in advance through experiments, etc. )
And exposure amount (exposure lamp, exposure laser, exposure LED
The control is performed by referring to the manipulated variables such as the drive voltage of the exposure light source and the pulse modulation width). A feedback control method is disclosed in, for example, Japanese Patent Laid-Open No.
As disclosed in Japanese Patent Laid-Open No. 3-151973, while varying such an operation amount, a device state is fed back by a sensor or the like to find an optimum operation amount by using a control method such as PID control. is there.

[0003]

However, in the case of the table reference method, the data to be referred to is limited,
It is difficult to accurately grasp and control the characteristics of the photoconductor.

Further, in the case of the feedback control method, the feedback loop is rotated many times until the ideal control state is reached, and the charging and exposure processes are performed each time, so that there is a lot of waste in time and the performance of the apparatus itself. (I.e. slow copy or print speed per hour, increased time to get first copy or print output)
Will be more likely to cause. Especially in the color electrophotographic process, since the potential corresponding to the intermediate density also needs to be controlled, the time required for control increases in proportion to the number of potentials to be controlled.

[0005]

According to a first aspect of the present invention, in an electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, a charging potential is set. , Exposed area potential,
Sensor means for acquiring information affecting the electrophotographic process such as temperature, humidity, total number of copies, photoconductor rotation time, photoconductor deterioration degree, and output data of the sensor means,
At least one of the parameters that affect the surface potential of the photoconductor such as the charging amount such as the charging charger voltage and the charging grid voltage, and the exposure voltage such as the driving voltage of the exposure light source and the pulse modulation width is input. , A photoreceptor formed by a neural network that has learned the latent image potential characteristics of the photoreceptor by using the relationship between the parameter of the photoreceptor and the surface potential of the photoreceptor obtained in advance as an instruction value. A means for estimating the surface potential of the photoconductor is provided.

According to a second aspect of the present invention, in an electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, a charging potential, an exposed portion potential, and a temperature. , Humidity, total number of copies, photoconductor rotation time, photoconductor deterioration degree, and other sensor means for acquiring information that affects the electrophotographic process, charge amount such as charging charger voltage and charging grid voltage, and exposure Output of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potentials of the photoconductors, such as the driving voltage of the light source for the light source, the exposure amount of the pulse modulation width, etc. Storage device for storing data, data such as charge amount and exposure amount sampled in time series, and output data of the sensor means sampled in time series And at least one of the parameters described above as an input, and using the relationship between the parameter of the photoconductor and the surface potential of the photoconductor obtained in advance based on experiments or the like as a teacher value, the latent image potential characteristic of the photoconductor. And a photoconductor surface potential estimating means for estimating the photoconductor surface potential formed by a learned neural network.

According to a third aspect of the present invention, in an electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, a charging potential, an exposed portion potential, and a temperature. , Humidity, total number of copies, photoconductor rotation time, photoconductor deterioration degree, and other sensor means for acquiring information that affects the electrophotographic process, charge amount such as charging charger voltage and charging grid voltage, and exposure Output of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potentials of the photoconductors, such as the driving voltage of the light source for light source, the amount of exposure such as pulse modulation width, etc. Storage device for storing data, data such as charge amount and exposure amount sampled in time series, and output data of the sensor means sampled in time series , An inclination γ obtained from a plurality of photoconductor surface potentials charged / exposed with a plurality of combinations of a charge amount and an exposure amount, and at least one of the parameters described above, which are obtained in advance based on experiments and the like. Photoconductor surface potential estimation for estimating the photoconductor surface potential formed by a neural network that has learned the latent image potential characteristics of the photoconductor using the relationship between the parameter of the photoconductor and the surface potential of the photoconductor as a teacher value. And means.

According to a fourth aspect of the present invention, in an electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, a charging potential, an exposed portion potential, and a temperature. , Humidity, total number of copies, photoconductor rotation time, photoconductor deterioration degree, and other sensor means for obtaining information that affects the electrophotographic process, and output data of this sensor means, charging charger voltage, charging grid voltage, etc. Of at least one of the parameters affecting the surface potential of the photoconductor such as the charge amount of the exposure light source, the driving voltage of the exposure light source, the exposure amount such as the pulse modulation width, etc. The latent image potential characteristic of the photoconductor is learned by using the relationship between the parameter of the photoconductor and the slope γ of the surface potential of the photoconductor as a teacher value. Provided a photosensitive member surface potential γ estimation means for estimating the inclination γ of the surface potential of the photosensitive member is formed by the workpiece.

According to a fifth aspect of the present invention, in an electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, a charging potential, an exposed portion potential, and a temperature. , Humidity, total number of copies, photoconductor rotation time, photoconductor deterioration degree, and other sensor means for acquiring information that affects the electrophotographic process, charging amount such as charging charger voltage, charging grid voltage, and exposure Output of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potentials of the photoconductors, such as the driving voltage of the light source for light source, the amount of exposure such as pulse modulation width, etc. Storage device for storing data, data such as charge amount and exposure amount sampled in time series, and output data of the sensor means sampled in time series As input at least one of the fine the parameter, the slope of the surface potential of the photosensitive member and the parameters of the photosensitive body obtained based on experiment or the like in advance γ
And a photoconductor surface potential γ estimation means for estimating the inclination γ of the photoconductor surface potential formed by a neural network that has learned the latent image potential characteristic of the photoconductor.

According to a sixth aspect of the present invention, in an electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, a charging potential, an exposed portion potential, and a temperature. , Humidity, total number of copies, photoconductor rotation time, photoconductor deterioration degree, and other sensor means for acquiring information that affects the electrophotographic process, charging amount such as charging charger voltage, charging grid voltage, and exposure Output of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potentials of the photoconductors, such as the driving voltage of the light source for light source, the amount of exposure such as pulse modulation width, etc. Storage device for storing data, data such as charge amount and exposure amount sampled in time series, and output data of the sensor means sampled in time series The sensitivity γ obtained from the surface potentials of a plurality of photoconductors charged / exposed with a plurality of combinations of the charge amount and the exposure amount, and at least one of the above parameters are input, and the photosensitivity obtained in advance based on an experiment or the like. Using the relationship between the parameter of the body and the slope γ of the surface potential of the photoconductor as a teacher value, the latent image potential characteristic of the photoconductor is formed by a learned neural network to estimate the slope γ of the photoconductor surface potential. A means for estimating the photoreceptor surface potential γ is provided.

According to the invention of claim 7, claims 1, 2,
With respect to the invention described in 4 or 5, the dither pattern is included in the parameters that affect the surface potential of the photoconductor.

According to the invention of claim 8, claim 3 or 6
Regarding the described invention, the dither pattern is included in the parameters that affect the surface potential of the photoconductor, and the slope γ
In addition to the charge amount and the exposure amount, a dither pattern is included in the plurality of combinations for obtaining.

[0013]

In the present invention, since the photoreceptor surface potential estimating means or the photoreceptor surface potential γ estimating means is formed by the neural network which learned the latent image potential characteristics of the photoreceptor, the generalization ability of the neural network is obtained. Due to
It is possible to perform highly accurate control in consideration of the characteristics of the photoconductor with less experiments. In other words, even if the number of potentials to be controlled is large, it is possible to find a combination of the charge amount, the exposure amount, etc., which can obtain the target potential in a short time. Therefore, the development period can be shortened and the cost can be significantly reduced as compared with the table reference method and the like.

Further, in the present invention as defined in claims 2 to 6, a neural network is used from the latent image potentials or a plurality of latent image potential gradients γ created by combining a plurality of charge amounts and exposure amounts. Since the latent image potential or the inclination thereof is estimated, the latent image potential can be more accurately estimated while monitoring the past history. In addition, there are long-term and short-term deterioration of the sensitivity of the photoconductor, but the next latent image potential is estimated from the latent image potential characteristics of the photoconductor from the past to the present. It is possible to detect and provide not only long-term photoreceptor fatigue but also short-term photoreceptor fatigue.

Further, in the inventions according to claims 4 to 6, since the inclination γ of the latent image potential obtained from a plurality of potentials is estimated instead of each one of the latent image potentials, one by one
The control time can be shortened as compared with the case of estimating and controlling one latent image potential.

Further, in the inventions of claims 7 and 8, since the dither pattern is taken into consideration in addition to the charge amount and the exposure amount as parameters, it is effective in the color electrophotographic process in which gradation is important. It becomes the target.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described with reference to FIGS. The present embodiment corresponds to the invention described in claim 1. First, the outline of the configuration of a copying machine which is an image forming apparatus to which the present invention is applied will be described with reference to FIG. This copying machine forms an image by an electrophotographic process such as charging, exposure, development, transfer and fixing. That is, the drum-shaped photoconductor 1 includes the charging operation value determination unit 2 (details will be described later).
Under the control of (1), after being uniformly charged by the charging charger 4 whose charge amount is controlled by the charging control unit 3, an electrostatic latent image is formed by being exposed by the exposure control unit (light source for exposure) 5. Here, in the exposure control unit 5, for example, an original 8 set on the contact glass 6 and covered with the original pressure plate 7 is exposed and illuminated by an exposure lamp 9, and reflected light from the original 8 is reflected by an optical system 10 such as a mirror. Through an image sensor, such as the CCD 11
After reading with, the A / D converter 12 and the document image processing unit 13
Further, the optical writing is performed with an exposure amount according to the document image information that has passed through the exposure operation value determination unit 14 (details will be described later), and includes an exposure laser, an exposure LED array, an exposure lamp, and the like. The electrostatic latent image thus formed on the photoconductor 1 is developed by the toner of the negative / positive developing type developing device 15 into a visible image, and the transfer sheet is fed at a predetermined timing in the transfer section 16. 17 is transcribed. The transferred transfer paper 17 is separated from the photoconductor 1 by the action of the separation unit 18, and then goes to the paper discharge side through the fixing roller 19 and the like.

As the sensor means, for example, a surface electrometer 20 for measuring the exposed portion potential on the surface of the photoconductor 1 after exposure, a temperature sensor for measuring environmental temperature and humidity around the photoconductor 1. 21, a humidity sensor 22, a copy counter (not shown) and the like are provided.

FIG. 4 shows potential relationships of the negative / positive developing system in the electrophotographic process of the copying machine as shown in FIG. First, the photoconductor 1 is charged to the charging potential V _D by the charging charger 4, and subsequently, a portion to be an image (exposure portion) is exposed by the exposure control portion 5 with an exposure amount that differs depending on the image density of the exposure portion potential _VL. Becomes Here, the developing bias voltage V _{B and} the exposed portion potential V
The difference from _L (V _B −V _L ) is called the development potential, and the amount of toner proportional to the potential difference which is the development potential adheres to the surface of the photoconductor 1 for development. In this way, the toner attached to the surface of the photoconductor 1 is transferred to the transfer paper 17 to form an image on the transfer paper 17. Therefore, it is important to control the potential of the exposed portion to an appropriate potential corresponding to the original density in order to obtain a high quality image in the electrophotographic process, particularly the color electrophotographic process.

Therefore, the relationship between the exposure amount (driving voltage, driving voltage for the exposure light source such as the exposure laser, the exposure laser, the exposure LED array, and the like) and the exposure portion potential will be examined. The relationship between the two is illustrated in FIG. In the color electrophotographic process, the slope γ and the curve of the exposed portion potential have to be appropriately controlled especially with respect to the writing level. This is because it is more important in the color electrophotographic process than in the monochrome electrophotographic process to faithfully match the halftone density to the original. That is, since the color electrophotographic process is a combination of toners of three colors or four colors, if the density of even one color is deviated, the color misregistration of the copy image is eventually caused.

Gradation is also important in the color electrophotographic process. For example, when the driving voltage of the exposure light source is 32 steps, the pulse modulation is 16 steps, and there are 8 types of dither patterns, among these 4096 combinations,
This is because the copy image greatly changes depending on which combination is selected and, for example, 256 gradations are obtained.

For example, with respect to the exposure characteristics shown in FIG. 5, the characteristics when the exposure amount with respect to the writing level is appropriately selected are as shown in FIGS. 6A and 6B. The characteristics when the exposure amount with respect to the level is improperly selected are as shown in FIGS. That is,
In FIG. 7 showing an inappropriate example, it is expected that the image density will hardly change at the writing level 4 or higher. As described above, even for the photoconductor 1 having the same exposure characteristic, an image faithful to the original cannot be obtained unless the exposure amount for the writing level is properly selected.

However, in this embodiment, the target potential is obtained in a short time even if the number of potentials to be controlled is large by performing the simulation by the neural network in which the latent image potential characteristic of the photoconductor 1 is learned. It is designed to find a combination of the charge amount, the exposure amount, and the like. For this reason, the charging operation value determination unit 2 and the exposure operation value determination unit 14 of the present embodiment.
Is composed of a neural network (photoconductor surface potential estimating means) 26 and a parameter operating section 27, as shown in FIG. Sensor (sensor means) 2
Reference numeral 8 denotes the surface electrometer 20, the temperature sensor 21, the humidity sensor 22 and the like described above.

The parameter operating unit 27 operates the parameters such as the charge amount and the exposure amount so that the output of the neural network 26 and the target value match. Here, in this embodiment, the surface potential of the photoconductor 1 is adjusted to the target value, and the target of the surface potential is determined for each writing level in order to obtain the ideal latent image potential. Then, the exposure amount is manipulated so that the estimated surface potential output from the neural network 26 matches the target value.
Also, the exposure amount when the estimated surface potential matches the target value is determined as the exposure amount for the writing level.

The neural network 26 uses each sensor 2
This is a network for estimating the surface potential of the photoconductor 1 based on the output of 8, and for example, a network that has been learned by the error back propagation method so that the error with the teacher value becomes small is used.
Here, this neural network 26 is shown in FIG.
It has a hierarchical structure as shown in, and learning is performed as follows.

In an experiment for acquiring learning data of the neural network 26, parameters such as charge amount (charge charger voltage, charge grid voltage, etc.), exposure amount (driving voltage of exposure light source such as exposure lamp, pulse modulation width, etc.), While changing the combination of environmental conditions (temperature, humidity, etc.) and charging potential,
The surface potential of the photoconductor 1 is sampled by copying with the photoconductor 1 having various sensitivities (total number of copies or different photoconductor rotation time) or by taking a pattern for potential measurement. In the neurons of the input layer of the neural network 26, the charge amount Vg, the exposure amount (driving voltage Ld, the pulse modulation width Pwm), the charge potential Vd, the environmental condition (temperature T, humidity H), the photoconductor fatigue degree Fc (total copy) By inputting one or more of the parameters such as the number of sheets, the total rotation time of the photoconductor, etc.) and giving the neuron of the output layer the surface potential of the exposure section as a teacher value, error back propagation Learn by law.

In the electrophotographic process control apparatus including the learned neural network 26, parameters that affect the latent image potential of the photoconductor 1 such as output data of each sensor 28, charge amount and exposure amount are controlled at the time of control. By inputting one or more of the above to the neural network 26, the surface potential of the photoreceptor 1 under that environment can be obtained. Then, the exposure amount is mainly manipulated so that the estimated surface potential Vp (t), which is the output of the neural network 26, becomes the target value, and the exposure amount when the estimated surface potential becomes equal to the target value is set to an arbitrary writing level. To determine the exposure amount.

As described above, according to this embodiment, the neural network 2 is used to estimate the surface potential (latent image potential) of the photoconductor.
By using 6, the latent image potential can be estimated by the generalization capability of the neural network 26 even for combinations of charge amount, exposure amount, environment, and photoconductor sensitivity for which latent image characteristics have not been acquired in the experiment. Therefore, it is possible to perform the control in consideration of the characteristics of the photoconductor 1 with less experiments. That is, the function can be realized with a smaller number of combinations of environmental factors and parameters such as the amount of charge and the amount of exposure. In other words, when trying to realize the same function by the table reference method, the accuracy of control is determined by the size of the table (the number of experiments), so more experiments are required when trying to perform highly accurate control. However, according to the present embodiment, the number of experiments required can be reduced by the generalization capability of the neural network 26.

Next, a second embodiment of the present invention will be described with reference to FIG. The same parts as those shown in the above-mentioned embodiments are designated by the same reference numerals (the same applies to the following embodiments).
This embodiment corresponds to the invention according to claim 7 based on the invention according to claim 1, and is basically similar to the first embodiment, but affects the surface potential of the photoconductor 1. The parameters to be exerted include at least a dither pattern in addition to the charge amount and the exposure amount. Therefore, as the neural network 29 of this embodiment, which replaces the neural network 26, the one learned by including the dither pattern Dth is used.

This learning is performed by the neural network 26.
It is almost the same as the learning of, but differs in that it includes a dither pattern. That is, the neural network 29 of this embodiment
In the experiment of learning data, the sensitivity such as charge amount, exposure amount, dither pattern (dither pattern is numbered for each pattern), environmental conditions, and combination of charge potential are changed, and various sensitivity is changed. By copying with the photoconductor 1 of or by taking the pattern for potential measurement,
The surface potential of the photoconductor 1 is sampled. The input layer neurons of the neural network 29 have a charge amount Vg
, Exposure amount, charging potential Vd, dither pattern Dth, environmental conditions (temperature T, humidity H), photoconductor fatigue degree Fc, and other parameters are input, and the neurons of the output layer are exposed to the surface of the exposed portion. By applying a potential as a teacher value, learning is performed by the error back propagation method.

In the electrophotographic process control device including the learned neural network 29 (in which the neural network 29 is used in place of the neural network 26 in the configuration of FIG. 1), the output of each sensor 28 at the time of control. Data and charge, exposure,
Dither pattern (number each dither pattern,
By inputting one or more of the parameters that affect the latent image potential of the photoconductor 1 such as inputting the number) to the neural network 29, the surface potential of the photoconductor 1 under the environment is can get. Then, the exposure amount and the dither pattern are mainly manipulated so that the estimated surface potential Vp (t), which is the output of the neural network 29, becomes the target value.
The exposure amount and dither pattern when the estimated surface potential becomes equal to the target value are determined as the exposure amount and dither pattern at an arbitrary writing level.

Therefore, the parameter operating unit 27 in this embodiment operates parameters such as the charge amount, the exposure amount, and the dither pattern so that the output of the neural network 29 and the target value match. Here, also in this embodiment, the surface potential of the photoconductor 1 is adjusted to the target value, and the target of the surface potential is determined for each writing level in order to obtain the ideal latent image potential. Then, the exposure amount and the dither pattern are manipulated so that the estimated surface potential output from the neural network 29 matches the target value. Further, the exposure amount and dither pattern when the estimated surface potential matches the target value are determined as the exposure amount and dither pattern for the writing level.

As described above, according to this embodiment, in addition to the first embodiment, since the neural network 29 also considers the dither pattern, the color electrophotographic process in which gradation is important is adopted. To be effective.

Next, a third embodiment of the present invention will be described with reference to FIGS. The present embodiment corresponds to the invention described in claim 2. The electrophotographic process control apparatus of this embodiment incorporates time series elements. Referring to FIG. 1, a memory device 30 is added, and instead of the neural network 26, a neural network (for surface potential estimation) is used. NN) 31 is used. Therefore, when the main part is rewritten, the structure is as shown in FIG. Here, the surface potential estimation NN31 receives as input the data obtained by sampling the output data of the sensor 28 in time series, and uses the latent image potential of the photoconductor 1 obtained in advance by an experiment or the like as a teacher value to determine the latent image potential of the photoconductor 1. The characteristics are previously learned as described later.

The learning of the surface potential estimating NN31 will be described. In an experiment in which the learning data of the surface potential estimation NN31 is used, parameters such as charge amount (charge charger voltage, charge grid voltage), exposure amount (drive voltage of exposure light source such as exposure lamp, pulse modulation width, etc.), environment By changing the combination of conditions (temperature, humidity, etc.) and charging potential, by taking a pattern for potential measurement on the photoconductors 1 with various sensitivities (total number of copies or photoconductor rotation time), the charging unit, The surface potential of the exposure section is sampled (n + 1) times (in this case, it is not always necessary to sample every copy, and for example, it may be sampled once every two or more copies). In this potential measuring pattern, a latent image is written at one or more writing levels, and as shown in FIG. 11, there are one or more latent image potential levels. Each writing level has a different exposure amount for each level. NN31 for surface potential estimation
To the neurons of the input layer of
Exposure amount (driving voltage Ld, pulse modulation width Pwm) at each writing level, latent image potential at each writing level, environmental conditions (temperature T, humidity H), charging potential Vd, and (n + 1) th charge amount. Input one or more of the exposure amount at the writing level, the environmental condition, and the charging potential, and use the latent image potential at the (n + 1) th writing level of the input layer in the pattern for potential measurement as a teacher value. By giving to the neurons of the output layer,
Learning by the error back propagation method. That is, unlike the learning of the above-described embodiment, the latent image potential characteristic of the photoconductor 1 is estimated from the potentials of the latent image potential patterns of the past n times.

The storage device 30 stores the output data from the sensor 28 and parameters such as the amount of charge and the amount of exposure. More specifically, the charging charger voltage
A plurality of photoconductors that are charged / exposed with a plurality of combinations of parameters that affect the surface potential of the photoconductor 1 such as the amount of charge such as the charging grid voltage and the amount of exposure such as the drive voltage of the light source for exposure and the pulse modulation width. Output data of the sensor 28 when the surface potential of the body 1 is measured, and parameters such as the amount of charge and the amount of exposure are sampled in time series as shown by t-n to t-1, t, and (n + 1) times of data are sampled. It is something to remember.

Such learned NN for surface potential estimation
In the electrophotographic process control device including 31, the output data of each sensor 28 sampled in time series and one of the parameters that affect the latent image potential of the photoconductor 1 such as the amount of charge and the amount of exposure at the time of control. Alternatively, by inputting a plurality of NN31 for surface potential estimation, the surface potential of the photoconductor 1 under the environment can be obtained. Then, the exposure amount is mainly manipulated so that the estimated surface potential Vp (t), which is the output of the surface potential estimation NN31, becomes the target value (only the data at time t), and the estimated surface potential becomes equal to the target value. The exposure amount at that time is determined as the exposure amount at an arbitrary writing level.

In this embodiment, in addition to the first embodiment described above, the sensitivity of the photosensitive member 1 is estimated from the potential of the latent image potential pattern in the past, and the next latent image potential is estimated.
It can be estimated more accurately. Further, the deterioration of the sensitivity of the photoconductor 1 may be caused by a long-term one (for example, a film thickness change caused by film scraping due to cleaning or the like) and a short-term one (for example, charging fatigue caused by repeated charging, exposure, and charge removal). However, in the present embodiment, the short-term deterioration cannot be detected by the method of determining the deterioration of the sensitivity of the photoconductor 1 from the total number of copies or the photoconductor rotation time. Since the next latent image potential is estimated from the transition of the latent image potential characteristic of the photoconductor 1 up to that point, not only long-term fatigue but also short-term fatigue can be detected.

Next, a fourth embodiment of the present invention will be described with reference to FIG.
13 and FIG. This embodiment corresponds to the invention according to claim 7 based on the invention according to claim 2, and is basically similar to the third embodiment, but is similar to the case of the second embodiment. In addition to the charge amount and the exposure amount, at least a dither pattern is included as a parameter that affects the surface potential of the photoconductor 1. Therefore,
As the surface potential estimating NN32 of this embodiment, which is an alternative to the surface potential estimating NN31, a learned one including the dither pattern Dth is used. In response to this, the storage device 30 is also supposed to store the past n dither patterns.

The learning of the surface potential estimating NN 32 will be described. This learning is almost the same as the learning of the surface potential estimating NN31 of the above-mentioned embodiment, but is different in that the dither pattern is included. In the experiment for acquiring the learning data of the surface potential estimating NN32, while changing the parameters such as the charge amount, the exposure amount, the dither pattern (the dither pattern numbers each pattern), the environmental conditions, and the charge potential, By taking a pattern for potential measurement with the photoconductor 1 having various sensitivities, the surface potentials of the charging part and the exposing part are (n + 1).
Sample once. This potential measurement pattern is 1
The latent image is written at three or more writing levels.
As in the case shown in, there will be one or more latent image potential levels. Each writing level is made up of a combination of an exposure amount and a dither pattern which are different for each level. In the neurons of the input layer of the surface potential estimation NN32, the charge amount Vg at the first to nth times, the exposure amount and dither pattern at each writing level, the latent image potential at each writing level,
Environmental conditions, charging potential, (n + 1) th charge amount, exposure amount and dither pattern at a certain writing level, environmental conditions,
By inputting one or more of the charging potentials and applying the latent image potential at the (n + 1) th write level of the input layer in the potential measurement pattern to the neuron of the output layer as a teacher value, the error Learn by the back propagation method. That is, similar to the learning of the above-described embodiment, the latent image potential characteristic of the photoconductor 1 is estimated from the potentials of the latent image potential patterns of the past n times.

Such learned NN for surface potential estimation
In the electrophotographic process control device including 32, at the time of control, among the parameters that influence the latent image potential of the photoconductor 1 such as the output data of each sensor 28 sampled in time series and the charge amount, the exposure amount, and the dither pattern. By inputting one or more of the above to NN32 for surface potential estimation,
The surface potential of the photoreceptor 1 under that environment can be obtained. Then, the exposure amount and the dither pattern are mainly operated so that the estimated surface potential Vp (t), which is the output of the surface potential estimation NN32, becomes the target value (only the data at the time t), and the estimated surface potential becomes the target value. The exposure amount and dither pattern when they are equal are determined as the exposure amount and dither pattern at an arbitrary writing level.

Therefore, the parameter operating unit 27 in the case of the present embodiment operates the parameters such as the charge amount, the exposure amount and the dither pattern so that the output of the surface potential estimating NN 32 and the target value match. Here, also in this embodiment, the surface potential of the photoconductor 1 is adjusted to the target value, and the target of the surface potential is determined for each writing level in order to obtain the ideal latent image potential. Then, the exposure amount and the dither pattern are manipulated so that the estimated surface potential, which is the output of the surface potential estimating NN 32, matches the target value. Further, the exposure amount and dither pattern when the estimated surface potential matches the target value are determined as the exposure amount and dither pattern for the writing level.

As described above, according to this embodiment, in addition to the third embodiment, since the surface potential estimating NN32 also considers the dither pattern, the color electrophotographic process in which gradation is important is performed. It will be effective for what you take.

Next, a fifth embodiment of the present invention will be described with reference to FIGS. This embodiment corresponds to the invention described in claim 3, and handles the data sampled in time series in the same manner as the third embodiment, and in addition to this, in addition to this, the slope γ of the surface potential of the photosensitive member 1 Is taken into consideration. Referring to FIG. 1, a storage device 30 is also added in the present embodiment, and a learned surface potential estimating NN 33 is used instead of the neural network 26. In response to this, the storage device 30 of this embodiment is supposed to store the data of the inclination γ as the surface potential data. Therefore, when the main part is rewritten, the structure is as shown in FIG. Here, the surface potential estimation NN 33 is the sensor 2 sampled in time series.
8 is used as input, at least one of the output data, the inclination γ obtained from the surface potentials of the plurality of photoconductors 1 charged / exposed by a plurality of combinations of the charge amount and the exposure amount, and the above parameters is input. The latent image potential characteristic of the photoconductor 1 has been learned by using the relationship between the parameter of the photoconductor 1 obtained based on the above and the surface potential of the photoconductor 1 as a teacher value.

The learning of the surface potential estimating NN 33 will be described. In the experiment for acquiring the learning data of the surface potential estimation NN33, parameters such as charge amount (charge charger voltage, charge grid voltage, etc.), exposure amount (drive voltage of exposure light source such as exposure lamp, pulse modulation width, etc.), environment By changing the combination of conditions (temperature, humidity, etc.) and charging potential, the pattern of potential measurement is taken by the photoconductors 1 with various sensitivities (total number of copies or photoconductor rotation time is different). The surface potential of the exposure section is sampled (n + 1) times (in this case, it is not always necessary to sample every copy, and for example, it may be sampled once every two or more copies). In this potential measuring pattern, a latent image is written at one or more writing levels, and as in the case shown in FIG. 11, one or more latent image potential levels exist. Each writing level has a different exposure amount for each level. For the neurons in the input layer of the surface potential estimation NN33, the charge amount Vg at the first to nth times, the exposure amount at each writing level (driving voltage Ld, pulse modulation width Pwm), and the potential gradient γ of the potential measurement pattern. , Environmental conditions (temperature T, humidity H), charging potential Vd, (n
+1) Input one or more of the amount of charge, the amount of exposure at a certain writing level, environmental conditions, and the charging potential, and at the (n + 1) th writing level of the input layer in the potential measurement pattern. By applying the latent image potential of as a teacher value to the neurons in the output layer, learning is performed by the back propagation method. That is, in the case of the learning of the above-described third embodiment, each level of the latent image potential measurement pattern is input one by one,
According to the learning of the present embodiment, the number of input data can be reduced by inputting the gradient γ of the potential of the pattern (a parameter representing the gradient obtained by the linear approximation or other method).

Such learned NN for estimating the surface potential
In the electrophotographic process control device including 33, one of the output data of each sensor 28 sampled in time series and the parameters that affect the latent image potential of the photoconductor 1 such as the charge amount and the exposure amount at the time of control. Alternatively, by inputting a plurality of NNs 33 for estimating the surface potential, the surface potential of the photoconductor 1 under the environment can be obtained. Then, the exposure amount is mainly manipulated so that the estimated surface potential Vp (t), which is the output of the surface potential estimation NN33, becomes the target value (only the data at time t), and the estimated surface potential becomes equal to the target value. The exposure amount at that time is determined as the exposure amount at an arbitrary writing level.

According to the present embodiment as described above, similarly to the third embodiment, the sensitivity of the photoconductor 1 is estimated from the latent image potential of the latent image potential pattern in the past, and the next latent image potential is estimated. So
Highly accurate estimation is possible, both long-term and short-term photoconductor fatigue can be detected, and in particular, the latent image potential gradient γ is learned as an input instead of the latent image potential. Therefore, the number of inputs can be reduced.

Next, a sixth embodiment of the present invention is shown in FIG.
17 and FIG. This embodiment corresponds to the invention according to claim 8 based on the invention according to claim 3, and is basically similar to the fifth embodiment, but in the case of the second and fourth embodiments. Similarly to, as parameters that affect the surface potential of the photoconductor 1, in addition to the charge amount and the exposure amount, at least the dither pattern is included, and the charge amount and the exposure amount are also included in a plurality of combinations for obtaining the inclination γ. In addition, a dither pattern is included. Therefore, N for surface potential estimation
As the surface potential estimation NN34 of this embodiment instead of N33, a learned one including the dither pattern Dth is used. In response to this, the storage device 30 is also supposed to store the past n dither patterns.

The learning of the surface potential estimating NN 34 will be described. This learning is almost the same as the learning of the surface potential estimating NN 33 of the above-mentioned embodiment, but is different in that the dither pattern is included. In the experiment for acquiring the learning data of the surface potential estimation NN34, while changing the parameters such as the charge amount, the exposure amount, the dither pattern (the dither patterns number each pattern), the environmental conditions, and the charge potential, By taking a pattern for potential measurement with the photoconductor 1 having various sensitivities, the surface potentials of the charging part and the exposing part are (n + 1).
Sample once. This potential measurement pattern is 2
The latent image is written at three or more writing levels.
As in the case shown in, there will be more than one latent image potential level. Each writing level is made up of a combination of an exposure amount and a dither pattern which are different for each level. For the neurons in the input layer of the surface potential estimation NN34, the charge amount Vg at the 1st to nth times, the exposure amount and dither pattern at each writing level, the potential gradient γ of the potential measurement pattern, the environmental conditions, the charging potential, (N + 1) th charge amount, exposure amount and dither pattern at a certain writing level,
Inputting one or more of environmental conditions and electrostatic potentials, and applying the latent image potential at the (n + 1) th writing level of the input layer in the pattern for potential measurement as a teacher value to the neuron in the output layer. Then, the error back propagation method is used for learning. That is, similar to the learning of the above-described embodiment, the latent image potential characteristic of the photoconductor 1 is estimated from the potentials of the latent image potential patterns of the past n times.

Such learned NN for surface potential estimation
In the electrophotographic process control apparatus including 34, during control, among the output data of each sensor 28 sampled in time series and the parameters that affect the latent image potential of the photoconductor 1 such as the charge amount, the exposure amount, and the dither pattern. By inputting one or more of the above to the surface potential estimation NN34,
The surface potential of the photoreceptor 1 under that environment can be obtained. Then, the exposure amount and the dither pattern are mainly operated so that the estimated surface potential Vp (t), which is the output of the surface potential estimation NN34, becomes the target value, and the exposure amount when the estimated surface potential becomes equal to the target value. And the dither pattern is determined as the exposure amount and the dither pattern at an arbitrary writing level.

Therefore, the parameter operating unit 27 in the present embodiment operates parameters such as the charge amount, the exposure amount and the dither pattern so that the output of the surface potential estimating NN 34 and the target value match. Here, also in this embodiment, the surface potential of the photoconductor 1 is adjusted to the target value, and the target of the surface potential is determined for each writing level in order to obtain the ideal latent image potential. Then, the exposure amount and the dither pattern are manipulated so that the estimated surface potential which is the output of the surface potential estimation NN34 matches the target value (only the data at time t). Further, the exposure amount and dither pattern when the estimated surface potential matches the target value are determined as the exposure amount and dither pattern for the writing level.

As described above, according to this embodiment, in addition to the fifth embodiment, since the surface potential estimating NN34 also considers the dither pattern, the color electrophotographic process in which gradation is important is performed. It will be effective for what you take.

Next, a seventh embodiment of the present invention will be described with reference to FIG. This embodiment corresponds to the invention described in claim 4 and is different from each of the above-described embodiments for estimating the latent image potential (surface potential), and in this embodiment, the neural network 35 for estimating the slope γ of the surface potential. Is used as a means for estimating the photoreceptor surface potential γ. For this reason, it may be more effective to control the slopes γ of a plurality of potentials corresponding to the low-density portion from the potential corresponding to the low-density portion, rather than controlling the individual potentials in a color electrophotographic process or the like. It focuses on the points. In such a case, it is considered that the time can be shortened by estimating the inclination γ of the latent image potential obtained by a plurality of combinations of the charge amount and the exposure amount, rather than estimating each potential. In particular, it is effective in shortening the control time when there are many potentials to be controlled in high gradation such as a color electrophotographic process.

This embodiment, which pays attention to such a point, is shown in FIG.
Referring to, the learned neural network 35 is used in place of the neural network 26. Here, the neural network 35
Is a parameter that affects the surface potential of the photoconductor 1 such as the output data of the sensor 28, the charge amount such as the charging charger voltage and the charging grid voltage, and the exposure voltage such as the driving voltage of the exposure light source and the pulse modulation width. The latent image potential characteristic of the photoconductor 1 is set by using at least one of Is already learned.

The learning of the neural network 35 will be described. In an experiment in which the learning data of the neural network 35 is taken, parameters such as charge amount (charge charger voltage, charge grid voltage, etc.), exposure amount (driving voltage of exposure light source such as an exposure lamp, pulse modulation width, etc.) and environmental conditions ( By changing the combination of temperature, humidity, etc.) and charging potential, by taking a pattern for potential measurement with the photoconductor 1 of various sensitivities (total number of copies or photoconductor rotation time is different), the charging unit, the exposure unit (N + 1) times the surface potential is sampled (in this case, it is not always necessary to sample each copy, and for example, it may be sampled once every two or more copies). In this potential measurement pattern, latent images are written at two or more writing levels, and as in the case shown in FIG. 11, there are two or more latent image potential levels. Each writing level has a different exposure amount for each level. In the neurons of the input layer of the neural network 35, the charge amount Vg, the exposure amount at each writing level (driving voltage Ld, pulse modulation width Pwm), environmental conditions (temperature T, humidity H), charging potential Vd, photoconductor fatigue Degree (a value obtained based on the total number of copies, the total rotation time of the photoconductor, etc.) and one or more of the environmental conditions, and the gradient γ of latent image potential (the gradient obtained by linear approximation or other method). , Which is a representative parameter of (), is given to the neurons in the output layer as a teacher value to learn by the error back-propagation method. That is, the latent image potential is set to 1 as in the case of the neural network 26 of the first embodiment described above.
Instead of estimating one by one, the inclination γ of the latent image potential is estimated at one time.

In the electrophotographic process control device including the learned neural network 35, the output data of each sensor 28 and the parameters that affect the latent image potential of the photoconductor 1 such as the charge amount and the exposure amount are controlled at the time of control. One or more of the above combinations are input to the neural network 35, and the slope γ of the surface potential of the photoconductor 1 generated by the plurality of combinations of the input exposure amounts under the environment is obtained. Then, the exposure amount is mainly manipulated so that the gradient γ of the estimated surface potential, which is the output of the neural network 35, becomes the target value, and the combination of the exposure amounts when the gradient γ of the estimated surface potential becomes equal to the target value is obtained. The exposure amount at each writing level is determined.

Therefore, the parameter operating unit 27 in the present embodiment operates parameters such as the charge amount, the exposure amount, and the dither pattern so that the output of the neural network 35 and the target value match. Here, in the present embodiment, the slope γ of the surface potential of the photoconductor 1 is adjusted to the target value, and there are a plurality of values so that the slope γ of the estimated surface potential output from the neural network 35 matches the target value. Manipulate the exposure amount for the writing level. Then, each exposure amount when the slope γ of the estimated surface potential matches the target value is determined as the exposure amount for each writing level.

According to this embodiment, since the latent image potential gradient γ obtained from a plurality of potentials is estimated instead of each latent image potential, each latent image potential is estimated. The control time can be shortened as compared with the above-mentioned control.

Next, the eighth embodiment of the present invention is shown in FIG.
Will be described. This embodiment corresponds to the invention according to claim 7 based on the invention according to claim 4, and is basically similar to the seventh embodiment, but in the case of the second and fourth embodiments. Similarly to, in addition to the charge amount and the exposure amount, at least a dither pattern is included as a parameter that affects the surface potential of the photoconductor 1. Therefore, as the neural network 36 of the present embodiment, which replaces the neural network 35, the one learned by including the dither pattern Dth is used.

The learning of the neural network 36 will be described. This learning is almost the same as the learning of the neural network 35 of the above-mentioned embodiment, but is different in that the dither pattern is included. Neural network 36
In the experiment of learning data, the exposure amount of various sensitivities was changed while changing the parameters such as charge amount, exposure amount, dither pattern (dither patterns are numbered for each pattern), environmental conditions, and charging potential. By taking a pattern for potential measurement on the body 1, the surface potentials of the charging part and the exposing part are sampled (n + 1) times. In this potential measurement pattern, latent images are written at two or more writing levels, and as in the case shown in FIG. 11, there are two or more latent image potential levels. Each writing level is made up of a combination of an exposure amount and a dither pattern which are different for each level. To the neurons of the input layer of the neural network 36, one or more of the charge amount Vg, the exposure amount and dither pattern at each writing level, the environmental condition, the charge potential, and the photoconductor fatigue level are input to input the latent image potential. The learning is performed by the error back-propagation method by giving the gradient γ of γ to the neuron in the output layer as a teacher value.

In the electrophotographic process control device including the learned neural network 36, the output data of each sensor 28 and the latent image potential of the photoconductor 1 such as the charge amount, the exposure amount and the dither pattern are influenced at the time of control. One or more of the combinations of the parameters that give the input to the neural network 36 are input, and the slope γ of the surface potential of the photoconductor 1 caused by the plurality of combinations of the input exposure amount and the dither pattern under the environment is obtained. To be Then, the exposure amount and the dither pattern are mainly manipulated so that the gradient γ of the estimated surface potential, which is the output of the neural network 36, becomes the target value, and the exposure amount when the gradient γ of the estimated surface potential becomes equal to the target value. And the dither pattern combination is determined as the exposure amount and the dither pattern at each writing level.

Therefore, the parameter operating unit 27 in this embodiment operates parameters such as the charge amount, the exposure amount, and the dither pattern so that the output of the neural network 36 and the target value match. Here, in the present embodiment, the inclination of the surface potential of the photoconductor 1 is adjusted to the target value, as in the previous embodiments, and the inclination γ of the estimated surface potential output from the neural network 36 matches the target value. The exposure amount and the dither pattern for a plurality of writing levels are manipulated as described above. Then, a combination of each exposure amount and the dither pattern when the estimated surface potential gradient γ matches the target value is determined as a combination of the exposure amount and the dither pattern for each writing level.

As described above, according to this embodiment, in addition to the seventh embodiment, since the neural network 36 also considers the dither pattern, the color electrophotographic process in which gradation is important is adopted. It will be effective.

Next, a ninth embodiment of the present invention will be described with reference to FIGS. The present embodiment corresponds to the invention described in claim 5. The electrophotographic process control apparatus of the present embodiment is one in which time-series elements are incorporated in the seventh embodiment. Referring to FIG. 1, a memory device 30 is added and a neural network 26 is replaced with a neural network. The network (NN for estimating surface potential γ) 37 is used. Therefore, when the main part is rewritten, the structure is as shown in FIG. The surface potential γ estimation NN 37 receives as input the data obtained by sampling the output data of the sensor 28 in time series, and uses the latent image potential of the photoconductor 1 previously obtained by an experiment or the like as a teacher value for the latent image of the photoconductor 1. The potential characteristic is previously learned as described later.

The learning of the surface potential γ estimation NN 37 will be described. In the experiment for acquiring the learning data of the surface potential γ estimation NN37, parameters such as the charge amount (charge charger voltage, charge grid voltage, etc.), exposure amount (driving voltage of the exposure light source such as an exposure lamp, pulse modulation width, etc.) While changing the combination of environmental conditions (temperature, humidity, etc.) and charging potential,
The surface potentials of the charging part and the exposing part are sampled (n + 1) times by taking a pattern for potential measurement with the photoconductors 1 of various sensitivities (total number of copies or different photoconductor rotation time) (in this case, not necessarily It is not necessary to sample every copy, for example, 1 for every two or more copies.
It may be possible to sample twice). In this potential measurement pattern, latent images are written at two or more writing levels, and as in the case shown in FIG. 11, there are two or more latent image potential levels. Each writing level has a different exposure amount for each level. The neurons in the input layer of the NN37 for surface potential γ estimation have 1 to n
Charge amount Vg at the first time, exposure amount at each writing level (driving voltage Ld, pulse modulation width Pwm), latent image potential at each writing level, environmental conditions (temperature T, humidity H), charging potential Vd,
Input one or more of the (n + 1) th charge amount, the exposure amount at a certain writing level, the environmental condition, and the charge potential,
Error backpropagation is performed by giving a gradient γ (a parameter representing the gradient obtained by a linear approximation or other method) of the latent image potential at the (n + 1) th writing level of the input layer to the neuron of the output layer as a teacher value. Learn by law. That is, unlike the learning of the seventh embodiment described above, the latent image potential characteristic of the photoconductor 1 is estimated from the potentials of the latent image potential patterns of the past n times. Further, in comparison with the third embodiment, instead of estimating the latent image potentials one by one, the inclination γ of the latent image potential is estimated at one time.

The storage device 30 of this embodiment stores the output data from the sensor 28 and the parameters such as the charge amount and the exposure amount, as in the case of the third embodiment. More specifically, a plurality of combinations of parameters that affect the surface potential of the photoconductor 1 such as a charging amount such as a charging charger voltage and a charging grid voltage, and an exposure amount such as a driving voltage of an exposure light source and a pulse modulation width. The output data of the sensor 28 when measuring the surface potentials of the plurality of photoconductors 1 charged / exposed by, and parameters such as the charge amount and the exposure amount are tn to t-.
1, t are sampled in time series (n +
1) The data for one time is stored.

Such learned N for surface potential γ estimation
In the electrophotographic process control apparatus including the N37, at the time of control, among the combination of the output data of each sensor 28 sampled in time series and the parameters that affect the latent image potential of the photoconductor 1 such as the charge amount and the exposure amount. By inputting one or more of them to the surface potential γ estimation NN 37, the slope γ of the surface potential of the photoconductor 1 under the environment can be obtained. Then, the exposure amount is mainly manipulated so that the slope γ of the estimated surface potential, which is the output of the surface potential γ estimation NN 37, reaches the target value (only the data at the time t), and the slope γ of the estimated surface potential.
Is determined as the exposure amount at each writing level.

In this embodiment, in addition to the seventh embodiment described above, the sensitivity of the photosensitive member 1 is estimated from the potential of the latent image potential pattern in the past, and the next latent image potential is estimated.
It can be estimated more accurately. Further, since the next latent image potential is estimated from the transition of the latent image potential characteristics of the photoconductor 1 from the past to the present, not only long-term fatigue but also short-term fatigue can be detected. Become.

Next, a tenth embodiment of the present invention will be described with reference to FIG.
And FIG. 23. This embodiment corresponds to the invention according to claim 7 based on the invention according to claim 5, and is basically similar to the ninth embodiment except that the second, fourth,
As in the case of the eighth embodiment, as a parameter affecting the surface potential of the photoconductor 1, at least a dither pattern is included in addition to the charge amount and the exposure amount.
Therefore, as the surface potential γ estimation NN38 of this embodiment, which is an alternative to the surface potential γ estimation NN37, a learned one including the dither pattern Dth is used. In response to this, the storage device 30 is also supposed to store the past n dither patterns.

Learning of the surface potential γ estimating NN 38 will be described. This learning is almost the same as the learning of the surface potential γ estimation NN 37 of the above-mentioned embodiment, but is different in that it includes a dither pattern. In the experiment in which the learning data of the surface potential γ estimation NN38 is taken, the charge amount, the exposure amount, and the dither pattern (dither patterns are numbered for each pattern).
While changing the combination of parameters such as the above, environmental conditions, and the charging potential, the surface potentials of the charging portion and the exposing portion are set to (n
+1) Sampling is performed once. This potential measurement pattern has a latent image written at two or more writing levels.
As in the case shown in FIG. 11, there will be two or more latent image potential levels. Each writing level is made up of a combination of an exposure amount and a dither pattern which are different for each level. For the neurons in the input layer of the surface potential γ estimation NN38, the charge amount Vg at the first to nth times, the exposure amount and dither pattern at each writing level, the potential gradient γ in the potential measurement pattern, the environmental conditions, the charging Input one or more of potential, (n + 1) th charge amount, exposure amount and dither pattern at each writing level, environmental conditions, and charge potential,
The gradient γ of the (n + 1) th latent image potential of the input layer is given as a teacher value to the neuron of the output layer to learn by the error back propagation method. That is, similarly to the learning of the ninth embodiment, the latent image potential characteristic of the photoconductor 1 is estimated from the potentials of the latent image potential patterns of the past n times.

Such learned N for surface potential γ estimation
In the electrophotographic process control device including the N38, at the time of control, the output data of each sensor 28 sampled in time series and a combination of parameters such as a charge amount, an exposure amount, and a dither pattern that affect the latent image potential of the photoconductor 1 are combined. By inputting one or more of them to the surface potential γ estimation NN 38, the slope γ of the surface potential of the photoconductor 1 under the environment can be obtained. Then, the exposure amount and the dither pattern are mainly manipulated so that the slope γ of the estimated surface potential, which is the output of the surface potential γ estimation NN 38, becomes the target value (only the data at time t), and the slope γ of the estimated surface potential is The combination of the exposure amount and the dither pattern when it becomes equal to the target value is determined as the combination of the exposure amount and the dither pattern at each writing level.

Therefore, the parameter operating unit 27 in this embodiment operates parameters such as the charge amount, the exposure amount, and the dither pattern so that the output of the NN 38 for estimating the surface potential γ and the target value match. . Here, also in the present embodiment, the inclination of the surface potential of the photoconductor 1 is adjusted to the target value, as in the above-described embodiment, and the surface potential γ estimation N
The exposure amount and the dither pattern for a plurality of writing levels are manipulated so that the slope γ of the estimated surface potential, which is the output of N38, matches the target value. Then, a combination of each exposure amount and the dither pattern when the estimated surface potential gradient γ matches the target value is determined as a combination of the exposure amount and the dither pattern for each writing level.

As described above, according to the present embodiment, in addition to the ninth embodiment, since the NN38 for estimating the surface potential γ also considers the dither pattern, the color electrophotographic process in which the gradation is important. It is effective for those that take.

Next, an eleventh embodiment of the present invention is shown in FIG.
And FIG. 25. This embodiment corresponds to the invention described in claim 6, handles the data sampled in time series in the same manner as the ninth embodiment, and in addition to this, in addition to this, the slope γ of the surface potential of the photoconductor 1 Is taken into consideration. Referring to FIG. 1, in the present embodiment as well, a storage device 30 is added, and a learned surface potential γ estimation NN 39 is used instead of the neural network 26. In response to this, the storage device 30 of this embodiment is supposed to store the data of the inclination γ as the surface potential data. Therefore, when the main part is rewritten, the structure is as shown in FIG. Here, the surface potential γ estimation NN 39 is an inclination γ obtained from the surface potentials of the plurality of photoconductors 1 charged / exposed by a plurality of combinations of the output data of the sensor 28 sampled in time series, the charge amount and the exposure amount. , And at least one of the above-mentioned parameters as an input, and the latent image potential of the photoconductor 1 is set as a teacher value by using the relationship between the parameter of the photoconductor 1 and the surface potential of the photoconductor 1 obtained based on experiments in advance. The characteristics have already been learned.

The learning of the surface potential γ estimating NN 39 will be described. In an experiment in which the learning data of the surface potential γ estimation NN39 is taken, parameters such as the charge amount (charge charger voltage, charge grid voltage, etc.), the exposure amount (driving voltage of the exposure light source such as the exposure lamp, the pulse modulation width, etc.), While changing the combination of environmental conditions (temperature, humidity, etc.) and charging potential,
The surface potentials of the charging part and the exposing part are sampled (n + 1) times by taking a pattern for potential measurement with the photoconductors 1 of various sensitivities (total number of copies or different photoconductor rotation time) (in this case, not necessarily It is not necessary to sample every copy, for example, 1 for every two or more copies.
It may be possible to sample twice). In this potential measurement pattern, latent images are written at two or more writing levels, and as in the case shown in FIG. 11, there are two or more latent image potential levels. Each writing level has a different exposure amount for each level. The neurons in the input layer of the NN39 for surface potential γ estimation have 1 to n
Of the charge amount Vg at the first time, the exposure amount at each writing level (driving voltage Ld, pulse modulation width Pwm), the potential gradient γ of the potential measurement pattern, the environmental conditions (temperature T, humidity H), and the charging potential Vd. Input one or more of (n + 1) in the input layer
Learning by the error back-propagation method is performed by giving a gradient γ (a parameter representing the gradient obtained by a linear approximation or other method) of the latent image potential to the neurons in the output layer as a teacher value. That is, if the learning of the ninth embodiment described above,
Although each level of the latent image potential measurement pattern is input one by one, according to the learning of the present embodiment, the number of input data can be reduced by inputting the potential gradient γ of the pattern. it can.

Such learned N for surface potential γ estimation
In the electrophotographic process control device including N39, at the time of control, a plurality of combinations of parameters that influence the latent image potential of the photoconductor 1 such as output data of each sensor 28 sampled in time series and charge amount and exposure amount are controlled. By inputting one or more of them to the NN39 for estimating the surface potential γ, the slope γ of the surface potential of the photoconductor 1 under the environment can be obtained. Then, the exposure amount is mainly manipulated so that the slope γ of the estimated surface potential, which is the output of the surface potential γ estimation NN 39, becomes the target value (only the data at the time t), and the slope γ of the estimated surface potential becomes the target value. The combination of the exposure amounts when they are equal is determined as the exposure amount at each writing level.

According to the present embodiment as described above, similarly to the ninth embodiment, the sensitivity of the photoconductor 1 is estimated from the latent image potential of the latent image potential pattern in the past, and the next latent image potential is estimated. So
Highly accurate estimation is possible, both long-term and short-term photoconductor fatigue can be detected, and in particular, the latent image potential gradient γ is learned as an input instead of the latent image potential. Therefore, the number of inputs can be reduced.

Next, a twelfth embodiment of the present invention is shown in FIG.
6 and FIG. 27. This embodiment corresponds to the invention according to claim 8 based on the invention according to claim 6, and is basically similar to the above eleventh embodiment, but is different from the case of the sixth embodiment. Similarly, as a parameter affecting the surface potential of the photoconductor 1, in addition to the charge amount and the exposure amount, at least a dither pattern is included, and the charge amount and the exposure amount are included in a plurality of combinations for obtaining the inclination γ. In addition to the above, a dither pattern is included. Therefore, the surface potential γ
NN for estimating the surface potential γ of the present embodiment instead of the NN39 for estimating
As for 40, what is learned by including the dither pattern Dth is used. Corresponding to this, the storage device 30 also has past n
It is supposed to store the dither pattern for each batch.

The learning of the surface potential γ estimation NN 40 will be described. This learning is almost the same as the learning of the surface potential γ estimation NN39 of the above-mentioned embodiment, but is different in that the dither pattern is included. In the experiment in which the learning data of the surface potential γ estimation NN40 is used, the charge amount, the exposure amount, and the dither pattern (dither patterns are numbered for each pattern).
While changing the combination of parameters such as the above, environmental conditions, and the charging potential, the surface potentials of the charging portion and the exposing portion are set to (n
+1) Sampling is performed once. This potential measurement pattern has a latent image written at two or more writing levels.
As in the case shown in FIG. 11, there will be two or more latent image potential levels. Each writing level is made up of a combination of an exposure amount and a dither pattern which are different for each level. For the neurons in the input layer of the surface potential γ estimation NN 40, the charge amount Vg at the first to nth times, the exposure amount and dither pattern at each writing level, the potential gradient γ of the potential measurement pattern, the environmental conditions, the charging potential , (N + 1) th charge amount, exposure amount and dither pattern at each writing level,
One or more of environmental conditions and charged potentials are input, and the gradient γ of the (n + 1) th latent image potential of the input layer is given as a teacher value to the neuron of the output layer to learn by the error back propagation method. .

Such learned N for surface potential γ estimation
In the electrophotographic process control device including the N40, at the time of control, the output data of each sensor 28 sampled in time series and a plurality of parameters that affect the latent image potential of the photoconductor 1 such as the charge amount, the exposure amount, and the dither pattern are controlled. By inputting one or a plurality of the combinations of 1 to the surface potential γ estimation NN 40, the slope γ of the surface potential of the photoconductor 1 under the environment can be obtained. Then, the surface potential γ estimation NN4
The exposure amount and the dither pattern are mainly manipulated so that the estimated surface potential gradient γ which is an output of 0 becomes the target value, and the exposure amount and the dither pattern when the estimated surface potential gradient γ becomes equal to the target value Is determined as the combination of the exposure amount and the dither pattern at each writing level.

As described above, according to this embodiment, in addition to the eleventh embodiment, the NN 40 for estimating the surface potential γ also considers the dither pattern, so that the gradation is important in the color electrophotography. It is effective for those who take the process.

[0082]

According to the present invention, since the photoconductor surface potential estimating means or the photoconductor surface potential γ estimating means formed by the neural network in which the latent image potential characteristic of the photoconductor is learned is provided, the neural network has Due to the generalization ability, it is possible to perform highly accurate control in consideration of the characteristics of the photoconductor with less experiments, and thus it is possible to shorten the development period for obtaining the same function as compared with the table reference method and the like. The cost can also be reduced significantly. In particular, according to the invention described in claims 2 to 6, the past latent image potentials sampled in time series, or the inclinations γ of the latent image potentials created by combining a plurality of charge amounts and exposure amounts.
Therefore, since the latent image potential or its inclination is estimated using the neural network, the latent image potential can be estimated more accurately while monitoring the past history.
In addition, since the next latent image potential is estimated from the latent image potential characteristics of the photoconductor from the past to the present, it is possible to detect not only long-term photoconductor fatigue but also short-term photoconductor fatigue. According to the invention of claims 4 to 6, the gradient γ of the latent image potential obtained from a plurality of potentials is estimated instead of the potential of each latent image. The control time can be shortened as compared with the case where each latent image potential is estimated and controlled. Further, according to the invention of claims 7 and 8, in addition to the charge amount and the exposure amount as parameters, Since the dither pattern is also taken into consideration, the dither pattern can be effectively used in the color electrophotographic process in which gradation is important.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a schematic diagram showing the configuration of the neural network.

FIG. 3 is a schematic configuration diagram showing the entire copying machine.

FIG. 4 is an explanatory diagram showing a potential relationship of a negative / positive developing system.

FIG. 5 is a characteristic diagram showing a relationship between a photosensitive member surface potential and an exposure amount.

FIG. 6 is a characteristic diagram when an exposure amount is appropriately selected for a writing level.

FIG. 7 is a characteristic diagram when the exposure amount is inappropriately selected with respect to the writing level.

FIG. 8 is a schematic diagram of a neural network configuration showing a second embodiment of the present invention.

FIG. 9 is a block diagram showing a third embodiment of the present invention.

FIG. 10 is a schematic diagram showing the configuration of the neural network.

FIG. 11 is an explanatory diagram showing an example of a pattern for potential measurement which is configured by a plurality of write levels.

FIG. 12 is a block diagram showing a fourth embodiment of the present invention.

FIG. 13 is a schematic diagram showing the configuration of the neural network.

FIG. 14 is a block diagram showing a fifth embodiment of the present invention.

FIG. 15 is a schematic diagram showing the configuration of the neural network.

FIG. 16 is a block diagram showing a sixth embodiment of the present invention.

FIG. 17 is a schematic diagram showing the configuration of the neural network.

FIG. 18 is a schematic diagram of a neural network configuration showing a seventh embodiment of the present invention.

FIG. 19 is a schematic diagram of a neural network configuration showing an eighth embodiment of the present invention.

FIG. 20 is a block diagram showing a ninth embodiment of the present invention.

FIG. 21 is a schematic diagram showing the configuration of the neural network.

FIG. 22 is a block diagram showing a tenth embodiment of the present invention.

FIG. 23 is a schematic diagram showing the configuration of the neural network.

FIG. 24 is a block diagram showing an eleventh embodiment of the present invention.

FIG. 25 is a schematic diagram showing the configuration of the neural network.

FIG. 26 is a block diagram showing a twelfth embodiment of the present invention.

FIG. 27 is a schematic diagram showing the configuration of the neural network.

[Explanation of symbols]

1 Photoconductor 26 Neural Network Forming Photoconductor Surface Potential Estimating Means 28 Sensor Means 29 Neural Network Forming Photoconductor Surface Potential Estimating Means 30 Storage Device 31-34 Neural Network Forming Photoconductor Surface Potential Estimating Means 35-40 Neural network forming a means for estimating photoreceptor surface potential γ

[Procedure amendment]

[Submission date] January 12, 1994

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0013

[Correction method] Change

[Correction content]

[0013]

In the present invention, since the photoreceptor surface potential estimating means or the photoreceptor surface potential γ estimating means is formed by the neural network which learned the latent image potential characteristics of the photoreceptor, the generalization ability of the neural network is obtained. Due to
It is possible to perform highly accurate control in consideration of the characteristics of the photoconductor with less experiments . Therefore, as compared with the table reference method, etc., together we can reduce development time, that as the Do cost may be greatly reduced. Also, the neural network
Since the simulation is performed by calculation using
The target potential can be obtained in a shorter time than the feedback control method.
That can find a combination of the amount of electrification, the amount of exposure, etc.
Becomes

[Procedure Amendment 2]

[Document name to be amended] Statement

[Name of item to be corrected] 0045

[Correction method] Change

[Correction content]

The learning of the surface potential estimating NN 33 will be described. In the experiment for acquiring the learning data of the surface potential estimation NN33, parameters such as charge amount (charge charger voltage, charge grid voltage, etc.), exposure amount (drive voltage of exposure light source such as exposure lamp, pulse modulation width, etc.), environment By changing the combination of conditions (temperature, humidity, etc.) and charging potential, the pattern of potential measurement is taken by the photoconductors 1 with various sensitivities (total number of copies or photoconductor rotation time is different). The surface potential of the exposure section is sampled (n + 1) times (in this case, it is not always necessary to sample every copy, and for example, it may be sampled once every two or more copies). In this potential measurement pattern, a latent image is written at two or more writing levels, and as in the case shown in FIG. 11, there are two or more latent image potential levels. Each writing level has a different exposure amount for each level. For the neurons in the input layer of the surface potential estimation NN33, the charge amount Vg at the first to nth times, the exposure amount at each writing level (driving voltage Ld, pulse modulation width Pwm), and the potential gradient γ of the potential measurement pattern. , Environmental conditions (temperature T, humidity H), charging potential Vd, (n
+1) Input one or more of the amount of charge, the amount of exposure at a certain writing level, environmental conditions, and the charging potential, and at the (n + 1) th writing level of the input layer in the potential measurement pattern. By applying the latent image potential of as a teacher value to the neurons in the output layer, learning is performed by the back propagation method. That is, in the case of the learning of the above-described third embodiment, each level of the latent image potential measurement pattern is input one by one,
According to the learning of the present embodiment, the number of input data can be reduced by inputting the gradient γ of the potential of the pattern (a parameter representing the gradient obtained by the linear approximation or other method).

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0055

[Correction method] Change

[Correction content]

The learning of the neural network 35 will be described. In an experiment in which the learning data of the neural network 35 is taken, parameters such as charge amount (charge charger voltage, charge grid voltage, etc.), exposure amount (driving voltage of exposure light source such as an exposure lamp, pulse modulation width, etc.) and environmental conditions ( By changing the combination of temperature, humidity, etc.) and charging potential while taking a pattern for potential measurement with the photoconductors 1 of various sensitivities (total number of copies or photoconductor rotation time is different), the charging unit, the exposure unit Measure the surface potential of. In this potential measurement pattern, latent images are written at two or more writing levels, and as in the case shown in FIG. 11, there are two or more latent image potential levels. Each writing level has a different exposure amount for each level. In the neurons of the input layer of the neural network 35, the charge amount Vg, the exposure amount at each writing level (driving voltage Ld,
Pulse modulation width Pwm), environmental conditions (temperature T, humidity H), charging potential Vd, photoconductor fatigue level (value obtained based on total number of copies, total photoconductor rotation time, etc.), one of environmental conditions Alternatively, by inputting a plurality of latent image potential gradients γ (parameters representing the gradients obtained by linear approximation or other methods) to the neurons in the output layer as a teacher value, learning is performed by the error back propagation method. That is, instead of estimating the latent image potentials one by one as in the case of the neural network 26 of the first embodiment described above, the gradient γ of the latent image potential is estimated at one time. .

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0060

[Correction method] Change

[Correction content]

The learning of the neural network 36 will be described. This learning is almost the same as the learning of the neural network 35 of the above-mentioned embodiment, but is different in that the dither pattern is included. Neural network 36
In the experiment of learning data, the exposure amount of various sensitivities was changed while changing the parameters such as charge amount, exposure amount, dither pattern (dither patterns are numbered for each pattern), environmental conditions, and charging potential. By taking a pattern for potential measurement on the body 1, the surface potential of the charged and exposed parts
To measure. In this potential measurement pattern, latent images are written at two or more writing levels, and as in the case shown in FIG. 11, there are two or more latent image potential levels. Each writing level is made up of a combination of an exposure amount and a dither pattern which are different for each level. The neurons in the input layer of the neural network 36 have a charge amount Vg, an exposure amount and a dither pattern at each writing level,
By inputting one or more of environmental conditions, charging potential, and photoreceptor fatigue, and applying the gradient γ of the latent image potential to the neurons in the output layer as a teacher value, learning is performed by the error back propagation method.

[Procedure Amendment 5]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 2

[Correction method] Change

[Correction content]

[Fig. 2]

[Procedure correction 6]

[Document name to be corrected] Drawing

[Correction target item name] Figure 8

[Correction method] Change

[Correction content]

[Figure 8]

[Procedure Amendment 7]

[Document name to be corrected] Drawing

[Name of item to be corrected] Fig. 11

[Correction method] Change

[Correction content]

FIG. 11

Claims (8)

[Claims] 1. An electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, wherein a charging potential, an exposure portion potential, a temperature, a humidity, and a total number of copies. Sensor means for acquiring information that influences the electrophotographic process such as photoconductor rotation time and photoconductor deterioration degree, and output data of the sensor means, charge amount such as charging charger voltage and charging grid voltage, and The drive voltage of the exposure light source, at least one of the parameters such as the exposure amount such as the pulse modulation width that affects the surface potential of the photoconductor is input, and The latent image potential characteristic of the photoconductor is formed by a learned neural network by using the relationship between the parameter and the surface potential of the photoconductor as a teacher value. Electrophotographic process control apparatus characterized by comprising a photosensitive member surface potential estimating means for estimating a surface potential. 2. An electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process for forming an electrostatic latent image on a photoconductor, wherein a charging potential, an exposure portion potential, a temperature, a humidity, and a total copy number. Sensor means for obtaining information that affects the electrophotographic process such as photoconductor rotation time and photoconductor deterioration degree, charging amount such as charging charger voltage and charging grid voltage, and driving voltage of the exposure light source, Output data, charge amount and exposure of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potential of the photoconductor such as exposure amount such as pulse modulation width A storage device for storing data obtained by time-sequentially sampling the amount and the like; and output data of the sensor means sampled in time-series and Of the latent image potential characteristic of the photoconductor, using as a teaching value the relationship between the parameter of the photoconductor and the surface potential of the photoconductor obtained based on an experiment or the like in advance. An electrophotographic process control device, comprising: a photoconductor surface potential estimating means which is formed by a network and estimates a photoconductor surface potential. 3. An electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, comprising: charging potential, exposure portion potential, temperature, humidity, total number of copies. Sensor means for obtaining information that affects the electrophotographic process such as photoconductor rotation time and photoconductor deterioration degree, charging amount such as charging charger voltage and charging grid voltage, and driving voltage of the exposure light source, Output data, charge amount and exposure of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potential of the photoconductor such as exposure amount such as pulse modulation width A storage device for storing data obtained by time-sequentially sampling the amount and the like, and a storage device for storing the output data of the sensor means, the charge amount, and the exposure amount sampled in time series. Inputting at least one of the inclination γ obtained from the surface potentials of a plurality of photoconductors charged and exposed by a combination of a number and the parameters, and the parameters of the photoconductor obtained based on experiments in advance. A photoconductor surface potential estimating means for estimating the photoconductor surface potential formed by a neural network that has learned the latent image potential characteristics of the photoconductor using the relationship with the surface potential of the photoconductor as a teacher value is provided. Characteristic electrophotographic process control device. 4. An electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, wherein a charging potential, an exposure portion potential, a temperature, a humidity, and a total number of copies. Sensor means for acquiring information that influences the electrophotographic process such as photoconductor rotation time and photoconductor deterioration degree, and output data of the sensor means, charge amount such as charging charger voltage and charging grid voltage, and The drive voltage of the exposure light source, at least one of the parameters such as the exposure amount such as the pulse modulation width that affects the surface potential of the photoconductor is input, and The latent image potential characteristic of the photoconductor is formed by a learned neural network using the relationship between the parameter and the slope γ of the surface potential of the photoconductor as a teacher value. Electrophotographic process control apparatus characterized by comprising a photosensitive member surface potential γ estimation means for estimating the inclination γ of the surface potential of the photosensitive member. 5. An electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process of forming an electrostatic latent image on a photoconductor, comprising: charging potential, exposure portion potential, temperature, humidity, total number of copies. Sensor means for obtaining information that affects the electrophotographic process such as photoconductor rotation time and photoconductor deterioration degree, charging amount such as charging charger voltage and charging grid voltage, and driving voltage of the exposure light source, Output data, charge amount and exposure of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potential of the photoconductor such as exposure amount such as pulse modulation width A storage device for storing data obtained by time-sequentially sampling the amount and the like, and an output data of the sensor means sampled in time-series and the parameters. Of the latent image potential characteristic of the photoconductor by using at least one of the parameters as a teacher value the relationship between the parameter of the photoconductor and the slope γ of the surface potential of the photoconductor obtained based on an experiment or the like. And a photoconductor surface potential γ estimation means for estimating the slope γ of the photoconductor surface potential formed by a completed neural network. 6. An electrophotographic process control apparatus for an image forming apparatus having an electrophotographic process mechanism including a process for forming an electrostatic latent image on a photoconductor, wherein a charging potential, an exposure portion potential, a temperature, a humidity, and a total number of copies. Sensor means for obtaining information that affects the electrophotographic process such as photoconductor rotation time and photoconductor deterioration degree, charging amount such as charging charger voltage and charging grid voltage, and driving voltage of the exposure light source, Output data, charge amount and exposure of the sensor means when measuring the surface potentials of a plurality of photoconductors charged / exposed by a plurality of combinations of parameters that affect the surface potential of the photoconductor such as exposure amount such as pulse modulation width A storage device for storing data obtained by time-sequentially sampling the amount and the like; Inputting at least one of the inclination γ obtained from the surface potentials of a plurality of photoconductors charged and exposed by a combination of a number and the parameters, and the parameters of the photoconductor obtained based on experiments in advance. Estimating the photoconductor surface potential γ, which is formed by a neural network that has already learned the latent image potential characteristics of the photoconductor by using the relationship with the photoconductor surface potential γ as the teaching value. And an electrophotographic process control device. 7. The electrophotographic process control apparatus according to claim 1, wherein the parameter affecting the surface potential of the photoconductor includes a dither pattern. 8. A dither pattern is included in the parameters that affect the surface potential of the photoconductor, and a dither pattern is included in addition to the charge amount and the exposure amount in a plurality of combinations for obtaining the inclination γ. 7. The electrophotographic process control device according to claim 3 or 6.