Explainable Artificial Intelligence (XAI) in Manufacturing: Methodology, Tools, and Applications

This book provides a comprehensive overview of the latest developments in Explainable AI (XAI) and its applications in manufacturing. It covers the various methods, tools, and technologies that are being used to make AI more understandable and communicable for factory workers. With the increasing use of AI in manufacturing, there is a growing need to address the limitations of advanced AI methods that are difficult to understand or explain to those without a background in AI. This book addresses this need by providing a systematic review of the latest research and advancements in XAI specifically tailored for the manufacturing industry. Artificial Intelligence (AI) are technologies that enable computers to imitate human behavior. The computing speed, storage capacity, reliability, and interconnectivity of computers combined with human reasoning patterns give AI the ability to solve complex and large-scale problems. Explainable Artificial Intelligence (XAI) is a new trend in AI. In the Chapter 1, XAI is first defined. A procedure for implementing XAI is also established. Then, through a literature analysis, applications of XAI in various domains, such as medicine, service, education, finance, health care, manufacturing, food, and military, are compared. Some representative cases in these domains are also reported. Subsequently, several applications of XAI in the manufacturing domain are reviewed, including explaining the classification process and result of jobs in a factory, explaining an ANN-based cycle time prediction method, comparing the effects of the components of an alloy using SHAP analysis, etc. Chapter 2, Applications of XAI for Forecasting in the Manufacturing Domain, focuses on forecasting, an important function of manufacturing systems. Many operation and production activities, such as cycle time forecasting, sales forecasting, unit cost reduction, predictive maintenance, yield learning, etc., are based on forecasting. This chapter takes job cycle time forecasting as an example. There are several applications of AI techniques for job cycle time prediction. Among these, ANN (or DNN) applications are the most effective, but very difficult for factory workers to understand or communicate. To address this issue, existing XAI techniques and tools for explaining the reasoning process and result of ANNs (or DNNs) are introduced. We first introduce XAI tools for visualizing operations in ANNs (or DNNs), such as ConvNetJS, TensorFlow, Seq2Seq, and MATLAB, and then mention XAI techniques for evaluating the effect, contribution, or importance of each input on the output, including partial derivation, odd ratio, out-of-bag (OOB) predictor importance, recursive feature elimination (RFE), and SHAP. Subsequently, XAI techniques for approximating the relationship between the inputs and output of an ANN (or DNN), especially simpler Machine Learning techniques like case-based reasoning (CBR), classification and regression tree (CART), RF, gradient boosted decision tree, eXtreme gradient boosting (XGBoost), and RF-based incremental interpretation, are introduced. The application of each XAI technique is supplemented by simple examples and corresponding MATLAB codes, allowing readers to learn quickly. Chapter 3, Applications of XAI for Decision Making in the Manufacturing Domain, deals with an important topic in factory management, namely improving the understandability of AI applications for group multi-criteria decision making in manufacturing systems. Decision making may be more critical to the competitiveness and sustainability of a manufacturing system than production planning and control because of its long-term and cross-functional impact. AI and Industry 4.0 technologies have many applications in this field, most of which can also be applied for other decision-making purposes in manufacturing systems. In the beginning, a systematic procedure is established for guiding the group multi-criteria decision-making process. Applications of AI and XAI in identifying targets are first reviewed. Subsequently, the applications of AI and XAI in selecting factors and developing criteria are presented. AI technologies are widely used to derive the priorities of criteria. Therefore, XAI techniques and tools for explaining such AI applications are particularly important. Chapter 4, Applications of XAI to Job Sequencing and Scheduling in Manufacturing, discusses a new field of applications of XAI in manufacturing—job sequencing and scheduling. It first breaks down job sequencing and scheduling into several steps and then mentions AI technologies applicable to some of these steps. It is worth noting that many AI applications are directed at the preparation of inputs required for scheduling tasks, rather than the processes of scheduling tasks, which is a distinctive feature of this field. Nevertheless, many AI technologies have been explained in other domains or fields. These explanations can provide a reference for explaining AI applications in job sequencing and scheduling. Therefore, some generic XAI techniques and tools for job sequencing and scheduling are reviewed, including: • Referring to the taxonomy of job scheduling problems; • Tailoring dispatching rule; • Textual description, pseudocode; • Decision tree, flowchart. In addition, job sequencing and scheduling problems are often formulated as mathematical programming (optimization) models to be optimized. AI technologies can be applied to find the optimal solutions to the models. Applications of GA are of particular interest because such applications are most common in job scheduling. Moreover, XAI techniques and tools for explaining GA can be easily extended to account for other evolutionary AI applications such as artificial bee colony (ABC), ant colony optimization (ACO), and PSO in job scheduling. Applicable XAI techniques and tools include: • Flowchart, textual description; • Chromosomal diagram; • Dynamic line chart, bar chart with baseline. Some novel XAI techniques and tools for explaining GA are also introduced: • Decision tree-based interpretation; • Dynamic transition and contribution diagram. The book includes real-world case studies and examples to illustrate the practical applications of XAI in manufacturing. It is a valuable resource for researchers, engineers, and practitioners working in the field of AI and manufacturing.