Perspective Chapter: AI-Driven Production Sequencing in the Era of Industry 4.0 and 5.0 – From Theory to Smart Manufacturing

1.1 AI in manufacturing: Basic principles

AI is about creating systems that can learn, reason, see patterns, and make decisions. These abilities grow together with computer science and engineering development [4]. In manufacturing, AI helps automate processes, improve efficiency, and predict maintenance before any problems happen.

Some main AI technologies in manufacturing are:

• Machine learning (ML): Uses past data to make predictions and support decisions.

• Deep learning (DL): Uses multi-layer neural networks to process images, signals, and control complex systems.

• Natural language processing (NLP): Extract useful information from manuals, maintenance logs, or customer feedback.

• Computer vision (CV): Helps automate the inspection of products and supports robots in collaborative tasks.

These technologies also work together with the Industrial Internet of Things (IIoT), cyber–physical systems (CPS), and big data analytics, which allow companies to monitor and optimise production constantly [5].

2.1 Model description

The LSJ model is an advanced method for scheduling production task initiation in dynamic demand environments. It integrates three main factors:

1. Customer priority – the importance and urgency of orders based on agreements and deadlines.

2. Product complexity – duration and resources required for production.

3. Resource availability – status of machines, equipment, and materials.

The LSJ is determined based on the significance of each customer order. This model calculates the significance value for each job that enters the production process and then arranges a sequence for launching products accordingly. In this scenario, each customer order represents a single job, consisting of a set of products with multiple operations. The number of machines is denoted by “m,” and the number of jobs by “n.” Typically, the subscript “i” denotes a job, while the subscript “j” refers to a machine. If a job requires multiple processing steps or operations, the pair (i, j) represents the specific step or operation of job “i” on machine “j.”

The problem can be described as follows: There exists a set of customer orders, {1, …, n}; these orders have been identified as jobs. For example, the first customer order will be Job1, the second customer order will be Job2, and so forth. Each order n consists of a set of products p with a set of operations o. The scheduling challenge is to determine the optimal LSJ by considering the value and significance of each customer order.

To address this, the model incorporates parameters related to business objectives, customer needs, and environmental conditions across the enterprise. These parameters are grouped into four main categories (or levels) that together influence the launch decision:

• Customer order significance (COS): urgency, importance, and profitability of the order.

• Product significance (PS): complexity, demand level, and value of products.

• Material significance (MS): availability and criticality of materials.

• Technological process significance (TPS): setup time, process stability, and machine requirements.

Overall, these factors define the LSJ model, which is represented by the following expression:

LSJ = f{COS, PS, MS, TPS}

In other words, the LSJ model prioritizes jobs by simultaneously evaluating customer priorities, product importance, material constraints, and technological requirements. The detailed mathematical formulation provided in the following subsections expands each of these four components into quantifiable sub-parameters.

While the mathematical formulation provides a rigorous representation of the LSJ model, the core logic can be summarized in simple terms. The LSJ evaluates each job according to several weighted criteria, such as customer importance, due date, technological setup time, and profitability. The resulting score determines the priority order. In practice, this means that high-value or urgent jobs are launched earlier, while low-impact or less time-sensitive jobs are delayed. The mathematical model formalizes this logic, but the intuition is that LSJ balances both operational efficiency and strategic business priorities.

According to our model created [6], the proposed model for launching sequences is as follows:

where:

Indices

i =1, 2, 3,…, n; i – customer order (jobs),

p =1, 2, 3, … pp; p – number of products,

o =1, 2, 3, …, oo; o – number of operations,

l =1, 2, 3, …, ll; l – number of parts.

Parameters

• x₁, x_{2, …,} x₇ – weighting factors determined through a digital logic model.

2.2 Operating mechanism

The LSJ model that is created can use real-time data from sensors and IIoT systems to update the job list. This integrated approach lets the system adapt dynamically to changing data, optimize maintenance planning, and reduce unplanned downtime. By combining it with reinforcement learning algorithms, the system can learn from past results and keep improving the job sequencing over time.

2.3 Impact on human roles

Contrary to fears of full worker replacement, LSJ places workers in new roles:

• Smart supervisors who monitor and adjust scheduling as needed.

• Strategic decision-makers who use data to optimize long-term processes.

• Collaborators with AI, combining practical experience with fast analyses from intelligent systems.

Practical example:

From a practical perspective, the model has been tested with data from a material processing company. Compared to other sequencing models, it turns out that this model is among the models that give the best results, and in special cases, when specific data of different natures must be obtained, it gives very good results. It should be noted that the model is not fully integrated with AI. Currently, some components, such as reinforcement learning for dynamic rescheduling, predictive maintenance inputs, and quality control feedback, are AI-driven. However, other aspects, including the initial weighting of criteria and final overrides, remain rule-based or rely on expert judgment. This hybrid structure reflects the current maturity of AI in manufacturing: while algorithms provide adaptability and real-time responsiveness, human supervision and rule-based logic ensure reliability and acceptance on the shop floor. The integration of the model with AI algorithms significantly affects the reduction of process changeover times and delays of urgent orders, thanks to the calculation of priority in real time.

3.1 Predictive and preventive maintenance

One of the most widespread AI applications in manufacturing is predictive maintenance, where machine sensors collect data on temperature, vibration, noise, and other operational parameters. ML algorithms analyse this data to predict faults before they occur, avoiding unplanned downtime [8].

Practical example:

Siemens uses a system called Sinalytics, which collects and analyses real-time data from industrial equipment, reducing maintenance costs by up to 30% and increasing machine uptime.

3.2 Production process optimisation

AI can analyse data from all production phases and suggest changes to optimize operation sequencing, reducing cycle time and material waste [9].

This is closely related to the LSJ model, which makes real-time decisions on the optimal production order.

Real example:

Toyota implemented an AI system to optimize assembly line scheduling, using historical and real-time data to minimize wait times between stations.

3.3 Quality control with CV

CV technologies, empowered by deep neural networks, can identify product defects with an accuracy often surpassing human visual inspection [10].

CV is a branch of AI that allows computers and systems to extract meaningful information from digital images or photos, videos, and other visual data – and then act on or make recommendations based on that information [11]. Furthermore, CV uses a large amount of data to perform analysis that leads to the distinction and subsequent recognition of images. To do this, CV uses one of the most effective types of ML algorithms, which is DL; more specifically, convolutional neural networks (CNNs). However, machine vision (MV) is a subset of CV, which refers to the study of methods and techniques by which artificial vision systems can be built and used usefully in practical applications. The introduction of MV in industrial processes is usually driven by the need to lower costs by improving efficiency (and, in turn, productivity), reduce errors (and improve quality), or collect more data. Just as important, it can also replace missing skilled workers or free employees from dangerous, hard, or tiring tasks in industry. This greatly increases quality consistency and reduces the number of returned products.

Real example:

General electric uses CV to inspect aircraft engine parts, achieving defect detection accuracy of over 99%.

3.4 Practical python code for sequencing optimisation

Below is a simplified Python example that uses a basic genetic algorithm to optimize production sequencing based on completion time and order priority.

import random

# Job list: (name, completion_time, priority)

jobs = [

(“Job_A”, 5, 3),

(“Job_B”, 8, 2),

(“Job_C”, 3, 5),

(“Job_D”, 6, 4)

]

# Fitness calculation function for a sequence

def calculate_fitness(sequence):

total_time = 0

total_score = 0

for job in sequence:

total_time+ = job[1]

total_score+ = job[2] * (1/total_time)

return total_score

# Creating initial population

population = [random.sample(jobs, len(jobs)) for _ in range(10)]

# Iterations for optimisation

for iteration in range(100):

# Sort by fitness

population.sort(key = lambda x: calculate_fitness(x), reverse = True)

# Select the best and generate a new generation

best = population[:5]

offspring = []

for parent in best:

copy_seq = parent[:]

random.shuffle(copy_seq)

offspring.append(copy_seq)

population = best + offspring

# Optimal result

population.sort(key = lambda x: calculate_fitness(x), reverse = True)

print(“Best sequence:”, population[0])

Explanation:

• Each job has a completion time and a particular priority.

• The genetic algorithm seeks a solution to achieve the best priority/time ratio.

• This is a basic illustration. In practical scenarios, data from the IIoT, as well as more sophisticated analytics, would be incorporated.

3.5 Sustainability features in AI-driven manufacturing

Sustainability is a very important and critical factor in Industry 5.0. Sustainability requires a balance between economic value and environmental and social value [12]. AI in job sequencing and production management develops a lower-cost model by cutting down energy costs, reducing waste, and enabling the optimal use of natural resources.

3.6 Energy consumption optimisation

According to Zhang et al., real-time monitoring of the work [13] adjusts machine operating hours or production schedules to off-peak hours by using AI, which reduces waste and can suggest modifications to decrease it.

For example:

Through the EcoStruxure AI platform, Schneider Electric automates the monitoring and optimisation of factory energy use, enabling savings of 20% on total annual energy bill expenditures.

3.7 Minimizing waste and unusable materials

AI technology plays a significant role in analyzing processes by identifying critical areas of waste and providing recommendations to tweak the designs or production parameters to minimize waste. This is particularly useful in additive manufacturing (three-dimensional [3D] printing) technologies, where AI-driven topological optimisation techniques can greatly decrease the amount of material used [14].

Practical example:

Siemens energy utilizes topological optimisation in its algorithms to enhance 3D-printed turbine components, achieving a 40% improvement in weight and material efficiency while also prolonging the life of the components.

3.8 Supply chain optimisation for sustainability

AI can analyse supply routes and select options that minimize CO₂ emissions, as well as predict delays to avoid urgent transport with high environmental impact [15].

Real example:

Unilever implemented AI systems to optimize distribution routes, reducing vehicle mileage by 15% and significantly lowering greenhouse gas emissions.

Link to LSJ Model

The LSJ model can integrate sustainability indicators into its decision-making algorithm, balancing production priorities with environmental goals. For example, a low-priority job with very high energy consumption could be scheduled during off-peak energy tariff periods, reducing costs and environmental impact.

4.1 Predicting supply chain disruptions

AI-driven predictive models allow companies to identify potential supply chain risks before they cause real problems [16]. AI combines data from suppliers, transporters, and markets to detect early signs of shortages or delays.

Real example:

IBM Watson Supply Chain uses AI algorithms to monitor global events, such as natural disasters or trade conflicts, alerting companies to take alternative measures.

4.2 Dynamic adaptation of production planning

AI can reconfigure production plans in real time to adapt to demand changes or resource shortages. Integrating this approach with the LSJ model allows for the immediate selection of the optimal job sequence based on the current situation.

Real example:

During the COVID-19 pandemic, Bosch used AI systems to dynamically change production priorities, focusing on high-demand items and minimizing line changeover times.

4.3 The role of employees in resilience

Industry 5.0 emphasizes the human role as an essential part of the production ecosystem. AI does not replace workers but empowers them with data-driven information, enabling faster and more accurate decisions in unforeseen situations [17].

In this context, employees act as:

• Coordinators between intelligent systems and practical operations.

• Decision-makers for crisis response strategies.

• Innovators in adapting work processes.

4.4 Link to long-term performance

Building a sustainable and crisis-resilient production system not only reduces short-term losses but also enhances brand reliability and ensures competitiveness in the global market. AI, combined with the LSJ approach, acts as a catalyst for this long-term performance.

5.1 Siemens – optimisation of electronic production lines

Siemens implemented an AI system based on reinforcement learning to manage assembly sequencing in electronic production lines. The system receives real-time data from IIoT sensors and uses it to determine the optimal production order.

Siemens – Electronics Works Amberg (EWA): Digital Twin and AI/Analytics for Production Optimisation

Technologies Used: Siemens’ Digital Enterprise platform, factory digital twin, IIoT sensors, and AI-driven analytics for planning, quality management, and production flexibility.

Reported results:

~1,000 product variants per day and 1 product per second in the same production space, enabled by full digitization of processes [18, 19].

Built-in quality of 99.9989%, achieved through full integration of design, engineering, and manufacturing data [18].

A 13-fold increase in production volume since 1990, without a significant increase in space or workforce, enabled by automation and continuous optimisation [19, 20].

High automation level in the value chain (including THT/SMT assembly; Through-Hole Technology (THT), Surface-Mount Technology (SMT)), with ~99.999% connection-level quality [21].

Operational benefits from the factory space digital twin: faster production line relocation planning, better space utilization, reduced travel costs, and the ability to conduct remote inspections and virtual tours [18].

5.2 BMW – AI-driven production and logistics optimisation

BMW developed a platform called Smart Production Scheduler, which uses AI algorithms to combine order priority, parts availability, and assembly station capacity.

Technologies used: AI-based orchestration of internal logistics, predictive maintenance systems, and full-scale digital twin simulations to optimize assembly line design and operation.

Reported results:

• Up to 30% reduction in internal material lead time at the Regensburg plant has been achieved through an AI-powered orchestration system for over 500 Automated Guided Vehicles (AGVs), which continuously share position and battery data for real-time route optimisation [22].

• Avoidance of approximately 500 minutes of assembly line disruption per year via AI-enabled predictive maintenance for conveyor systems, allowing faults to be anticipated and resolved before they cause downtime [23].

• Increased flexibility and customization capability through the use of a virtual factory and digital twin technology, based on NVIDIA Omniverse, enabling real-time simulation of assembly workflows and ergonomic adjustments before physical implementation [24, 25].

5.3 Bosch – supply crisis response

During the COVID-19 pandemic, Bosch used AI systems to analyse different supply scenarios and prioritize production according to parts availability.

Bosch – AI-driven supply chain resilience & supply crisis response

Technologies Used: AI, IoT, real-time digital supply chain platforms (including demand forecasting, inventory optimisation, procurement systems, digital twins, and end-to-end visibility tools).

Reported results:

• Up to 30% improvement in supply chain efficiency achieved through Bosch’s digital transformation initiatives, including AI-powered demand forecasting, predictive analytics, IoT tracking, procurement optimisation, and digital twin capabilities [26].

• Reduction in supply–demand mismatches by up to 35%, thanks to automated demand forecasting tools processing millions of data points per day [26].

• Enhanced supply chain resilience, enabled by real-time visibility via IoT and blockchain (e.g., the Trac360 platform), supports rapid scenario modeling, disruption response, and decentralized planning capabilities making agile responses possible within hours rather than days [26].