Smart manufacturing is the core of the Industry 4.0 (I4.0) concept, which has immensely increased the importance of production planning and control (PPC) in I4.0 activities. High computing power, copious availability of data, and large storage capacity with the advent of I4.0 have made artificial intelligence (AI) techniques, like machine learning (ML), an appealing solution for various functions in the manufacturing domain. This article deliberates on the importance and applications of AI/ML techniques for decision-intensive tasks in PPC and recent developments.
Importance of AI in PPC
PPC is a fundamental function of all manufacturing systems, implying extensive decision-making processes about the delivery plans, production plans, profitability, and productivity. The PPC function involves enterprise resource planning (ERP), material requirements planning (MRP), replenishment, forecasting, collaborative planning, and just-in-time manufacturing.
This is a value-adding process of the manufacturing activity that needs to be continually adapted to new supply chain opportunities, complex customer requirements, and operational and strategic environments. Thus, PPC must be integrative, adaptive, and dynamic. ML techniques reduce costs while improving the sustainability, quality, and productivity of the production process by leveraging the data collected during the production lifecycle. Specifically, the use of AI/ML can improve the decision-intensive functions in PPC.
Support vector machine (SVM), neural networks (NN), fuzzy logic, deep learning (DL), decision trees (DT), ant colony optimization (ACO), genetic algorithms (GA), deep neural networks (DNN), particle swarm optimization (PSO), regression analysis, fuzzy neural networks, pattern recognition, and random forest (RF) are the common AI/ML methodologies used in decision-intensive tasks within PPC.
Among them, SVM and NN are the most extensively utilized methods, followed by DL and DNN, which indicate the persistent utilization of artificial neural networks (ANN). The application of fuzzy logic and DT displayed the preference towards black box ML techniques and expert systems.
For instance, linear regression, logistic regression, DT, and NN are utilized for forecasting incoming orders, detection of customers with a high rate of changing orders, detection of dependencies between product quality and machine settings, and forecast of effects of unexpected events like machine failure or incoming rush orders, respectively. ML techniques also allow the comparison of planned and actual data and assist the decision-maker in discovering previously unknown connections between products or processes, workflow patterns, or deviations from the desired state.
In production optimization, ML algorithms are typically applied to improve the production line output quality, while DL and reinforcement learning (RL) approaches are used extensively in scheduling. Moreover, quality control is achieved in the microelectronic industry by the implementation of computer vision.
Similarly, a DL-based methodology can be used to calculate and evaluate the outcomes of various information technology and marketing strategies on the company’s operational performance. Dynamic production scheduling, performance evaluation and monitoring, and process automation and control are the key AI/ML applications in PPC-decision-intensive tasks.
Dynamic Scheduling
Smart scheduling is one of the most critical applications of AI/ML in PPC. The dynamic nature of the manufacturing systems as a result of their gradual transition towards seamless mass customization of products and production processes due to the advent of I4.0 has necessitated the adoption of a dynamic scheduling paradigm to effectively deal with unforeseen events that cause disruptions to the execution of a schedule. In dynamic scheduling, the assigned apparitions are immediately redirected to other machines.
Dynamic scheduling is task-oriented, line-oriented, or line-oriented in PPC. Tasks’ rescheduling involves rescheduling a specific task in the production process in response to a disruption in the original task schedule. An ANN-enhanced evolutionary algorithm is effective for solving the issue of simultaneous job processing on a particular machine.
Resource allocation deals with the resource-focused application of dynamic scheduling in flexible shop floors. The use of AI/ML in a dynamic production environment improves the ability to assign necessary resources to face disruptions in plans. For instance, a decision support system based on fuzzy analytic network process (FANP) and ANN was developed for resource allocation to improve the machine selection process in the context of flexible manufacturing.
Line balancing applications are related to production line balancing after possible disruptions in the production process. For instance, an SVM-based approach has been proposed for developing a self-healing process to predict process anomalies and perform dynamic reconfigurations for production line parameters.
Performance Evaluation and Monitoring
The monitoring and evaluating various production/manufacturing performance metrics is a fundamental task in the PPC of manufacturing systems. Such monitoring and evaluation allow identification of bottlenecks and facilitate decision-making during production operations.
Industrial data obtained using cloud manufacturing, the Internet of Things (IoT), and smart sensors are used extensively for this application. In PPC, two fundamental performance evaluation and monitoring aspects include performance prediction, optimization, and process fault detection.
Performance prediction and optimization are concerned with improving production metrics based on industrial data and historical production instances. Real-time data analytics collected from processes and machines allow the detection of manufacturing problems and constraints considering the overall industrial ecosystem.
Additionally, greater adaptability is achieved by leveraging AI/ML capabilities, resulting in higher flexibility and improved performance in performance prediction. For instance, a hybrid approach combining conditional inference trees with dimensionality reduction using principal component analysis (PCA) and regression trees has been developed for predicting the performance of a deep drilling process under various cooling and cutting conditions.
Process fault detection involves maintaining the overall quality of the processes and avoiding faults, thus a critical element of PPC. Implementing AI/ML-based approaches effectively increases the overall manufacturing system performance through accurate prediction and avoidance of production failures and anomalies. For instance, an SVM-based approach has been proposed to develop an early warning system to identify potential problems in egg production.
Process Automation
The ability of AI/ML to automate several processes and tasks in the manufacturing industry is one of the biggest advantages of these technologies that serve as a vital pillar of the fourth industrial revolution. In the context of production planning, automation implies production strategy automation to realize the production goals. Process automation is also concerned with automated control, which involves automating the manufacturing process to attain greater overall flexibility and adaptability.
Industrial process automation is driven by the need for scalable demand to ensure a dynamic production environment in manufacturing facilities, specifically focusing on higher manufacturing flexibility and customizability. Automated planning/AI planning is the centerpiece of the I4.0 automation paradigm as it addresses the need for sufficient strategies to realize a specific goal automatically. Thus, AI planning provides the opportunity to achieve a significant extent of manufacturing autonomy, which is a key I4.0 objective.
In the manufacturing industry, process control often lacks adaptability and flexibility, which hinders the mass customization of products and processes within the I4.0 framework. The application of AI/ML that constantly leverages industrial data leads to higher flexibility through automated control of different process parameters within the manufacturing system.
Recent Developments
A paper recently published in the Journal of Intelligent Manufacturing presented a new approach to solve a closed-loop supply chain (CLSC) management problem through a fuzzy logic-based decision-making system built on ML. This proposed system provides decisions to operate a production facility integrated into a CLSC to attain the production goals in the presence of uncertainties. The proposed approach can reject the effects of imbalances in the rest of the chain on the inventories of finished products and raw materials.
An intelligent algorithm was used for the plant operation supervision and task reprogramming to ensure the realization of process goals. ML techniques and fuzzy logic are combined to design the tool. Researchers investigated the method's effectiveness on an industrial hospital laundry, and satisfactory results were obtained, demonstrating the potential of incorporating this proposal into the I4.0 framework.
In conclusion, AI/ML algorithms improve decision-making in PPC tasks like dynamic scheduling, performance evaluation, and process automation, leading to a more adaptive and efficient manufacturing system. However, challenges like high cost, data dependency, and lack of transparency must be addressed to safely and efficiently exploit the abilities of AI.
References and Further Reading
Elbasheer, M., Longo, F., Nicoletti, L., Padovano, A., Solina, V., Vetrano, M. (2021). Applications of ML/AI for Decision-Intensive Tasks in Production Planning and Control. Procedia Computer Science, 200, 1903-1912. https://doi.org/10.1016/j.procs.2022.01.391
Franke, F., Franke, S., Riedel, R. (2021). AI-based Improvement of Decision-makers’ Knowledge in Production Planning and Control. IFAC-PapersOnLine, 55(10), 2240-2245. https://doi.org/10.1016/j.ifacol.2022.10.041
González Rodríguez, G., Gonzalez-Cava, J.M., Méndez Pérez, J.A. (2020). An intelligent decision support system for production planning based on machine learning. Journal of Intelligent Manufacturing, 31, 1257–1273. https://doi.org/10.1007/s10845-019-01510-y
Bueno, A., Godinho Filho, M., Frank, A. G. (2020). Smart production planning and control in the Industry 4.0 context: A systematic literature review. Computers & Industrial Engineering, 149, 106774. https://doi.org/10.1016/j.cie.2020.106774