Artificial intelligence (AI) techniques have been behind disruptive innovations in every industry. Based on AI techniques, large amounts of data can be converted into actionable insights and predictions. Manufacturers have frequently faced different kinds of challenges, such as unexpected machinery failures or defective product deliveries. Still, the adoption of AI techniques is expected to improve operational efficiency, enable the launch of new products, customize product designs, and plan future financial actions. Recently, manufacturers have been using AI techniques to improve the quality of their products, achieve greater speed and visibility across supply chains, and optimize inventory management.

Given that the attention and interest in AI techniques has been growing rapidly, it is time that the current state of the art of their practical applications be presented. The main aim of this special issue is to bring together the latest AI research and practical case studies of AI techniques in production engineering.

This special issue features 10 papers related to not only operation automation but also sophisticated skill transfer in manufacturers. Their subjects cover various advancements, such as failure diagnosis, product estimation, process planning, operation planning, and workpiece fixturing in the area of machining. Moreover, the authors boldly strive to apply AI technologies even to complex systems in manufacturing fields such as laser-assisted incremental forming, injection molded direct joining, and parts assembling.

We thank the authors for their interesting papers submitted for this special issue, and we are sure that both general readers and specialists will find the information the authors provide both interesting and informative. Moreover, we deeply appreciate the reviewers for their incisive efforts. Without these contributions, this special issue would not have been possible. We truly hope that this special issue triggers further research on AI techniques in production engineering.

The demand for the mass production of micro-lens arrays (MLAs) is increasing. An MLA is fabricated through an injection molding process, and its mold is manufactured by a five-axis high-precision machine tool using a small diameter endmill. A visual examination is not available to judge the quality of the mold while machining. Therefore, an effective process monitoring technology must be developed. A promising approach is to apply a servomotor current to in-process monitoring because as long as the servomotor works well, no external sensors, capital investment, or maintenance processes are required. From this perspective, a machine learning-based shape error estimation method using only the servomotor current is proposed. To explore the relationship between the motor current generated during micro-milling and the shape error of the mold, the servomotor current in X-, Y-, and Z-axes was recorded, and the corresponding shape error of the MLA mold was measured after machining. Input data were prepared by converting time-domain servomotor current data to frequency-domain data using short-time Fourier transform and reducing the dimensions of the data via principal component analysis. In terms of a meaningful label for the output data, the average shape error in the machined area corresponding to each window was provided. The input/output relationships were used to train five different machine learning models, and the accuracy of shape error estimation using each model was evaluated. In addition, the estimation accuracies using the X-, Y-, and Z-axes were compared to find the axis that senses the shape error with the highest accuracy. The results show that the non-linear method using the X-axis servomotor current information closest to the machining point achieved the highest shape error estimation accuracy.

As a result of the development of network technologies, diagnosis techniques that can collect machine states continuously and prognostic health management (PHM) are available in the factory. PHM technology is also beginning to be implemented in the machine tool field. However, few studies have described causality between feature values, including vibration and acoustic emission data, collected by machine and physical phenomena of failures under the actual use of machine tools. In the present paper, a PHM system of lubrication failure of bearings in CNC lathe spindles is developed. An acceleration sensor is used to collect machine states, and statistical feature parameters that characterize the lubrication failure are extracted from the obtained vibration data. Moreover, in order to clarify the cause-effect relation between the extracted feature parameters and physical phenomena of lubrication failure, several analyses using surface roughness measurement, residual stress measurement, and grease consistency measurement are conducted.

At present, machining with numerically controlled (NC) machine tools is mostly performed by NC programs generated by computer-aided design and computer-aided manufacturing (CAD/CAM) systems. However, even if the machining shape to be machined is the same, there are numerous machining processes involving a series of operations such as determining the machining area, machining order, and machining conditions. These are entrusted to the user, and automation is difficult. In addition, these tasks depend on the experience and know-how of skilled engineers, and it is very difficult to convert them into algorithms and reflect them in the creation of NC programs. Therefore, in this study, artificial intelligence (AI) was used for the process design of multi-tasking machine tools, with the goal of determining and automating the process design using shape examples. We propose a shape recognition method that includes image analysis by AI. This image analysis makes it possible to determine the characteristics of the machining shape, and the machining operator can easily judge the machining process based on the CAD model. Furthermore, because there are shapes that cannot be determined from image data alone, shape features are also extracted from the STEP file of the CAD model. A language analysis of the STEP file can find the characteristic components and their numerical information to determine the coordinates of the shape features. By combining image analysis and language analysis, the method can easily judge the process based on the information in the CAD model. Finally, using the generated learning model and analysis program, we conducted a test to determine whether a multitasking machine tool is necessary for machining.

Process planning is well known as the key toward achieving highly efficient machining. However, it is difficult to standardize machining skills for process planning, which depend heavily on skilled operators. Hence, in a previous study, a computer aided process planning (CAPP) system using machine learning is developed to determine the operation parameters for finish machining of dies and molds. On the other hand, in rough machining, it is assumed that some machining operations are conducted sequentially using a respective tool according to the workpiece shape, which induces a much higher complexity in process planning. Therefore, in this study, machine learning is adopted to determine operation parameters for rough machining. The developed CAPP system converts the removal volume into a voxel model and infers a machining operation for each voxel. The inferred machining operation is visualized using different colors and identified corresponding to the voxel. Finally, the removal volume is classified using three different machining operations. However, machine learning is said to have a critical problem in that it is difficult to understand the reasons for the inferred results. Hence, it is necessary for the CAPP system to demonstrate the certainty level of the determined operation parameters. Thus, this study proposes a method for calculating the degree of certainty. If an artificial neural network is trained sufficiently, similar inferred results would always be obtained. Consequently, by using the Monte Carlo dropout to delete weights at random, the certainty level is defined as the variance of the inferred results. To verify the usefulness of the CAPP system, a case study is conducted by assuming rough machining of dies and molds. The results confirm that the machining operations are inferred with high accuracy, and the proposed method is effective for evaluating the certainty of the inferred results.

In parts machining, process planning is typically conducted by skillful operators. The quality of machining is highly dependent on process planning, which determines the operation parameters, such as the operation sequence and cutting tool. To achieve high-quality machining without depending on the skill level of the operators, standardization of process planning is desired. Therefore, it is necessary to extract and generalize skills related to process planning. Furthermore, eye tracking technology is expected to visualize unconscious human behavior. In this study, eye tracking technology is adopted to detect the movement of the operator’s eyes and gather gaze data when understanding mechanical drawings. Gaze data are analyzed using a heat map and bubble chart to identify differences in eye movement according to skill level. The analyzed heat maps indicate that the gazes of the skillful operator are gathered because the operator focuses on the area that is strongly related to the quality of machining. The analyzed bubble charts also indicate that the skillful operator considers the machining process by checking annotations, then understands the shape, and finally verifies the numerical values of the annotations. From the results of interviews performed based on the analysis, the individual skill could be effectively extracted in detail, particularly the skill regarding the operation sequence. Furthermore, the acquired skills are incorporated into a computer-aided process planning system developed in a previous study. The operation sequence is modified to reflect the acquired skills. Machining experiments confirmed the effectiveness of adopting operators’ skills in process planning.

For machining operations, preparation work called a “setting operation” is always required in advance. The setting operation, which affects the lead time and machining accuracy, strongly depends on the skill level of the operator. Therefore, to improve the quality of machining operations, skill transfer is necessary by extracting and generalizing the skills related to the setting operation. In addition, a variety of accidents often occur during the setting operation. This can lead to machine tool failure or a serious incident involving the operator. Thus, skill transfer to an unskilled operator is also important for work safety. On the other hand, augmented reality (AR) is a promising human-computer interaction technology to support skill transfer at the manufacturing site. An AR technology generally overlays virtual images on the real-world environment. In this study, an AR-based system is developed to demonstrate a recommended workpiece fixturing method in turning operations for assisting unskilled operators as the first step of skill transfer. In turning operations, two types of fixturing are usually assumed: outer diameter clamping and inner diameter clamping. The dimensions of the targeted product shape are detected, and the workpiece shape is obtained. The removal volume to be machined is calculated as the difference between the targeted product shape and workpiece shape. The fixturing method is determined to avoid contact between the removal volume and the assumed jaw. The results of a case study show that the developed AR-based system is effective for skill transfer of workpiece fixturing by demonstrating the recommended fixturing method using skills acquired from operators.

A three-dimensional (3D) printer can be used to form various shapes in a single process. However, shell shape formation is difficult because of the low adhesion strength between the layers in 3D printing, and sufficient stiffness cannot be maintained. Therefore, the authors focused on laser-assisted incremental forming, which enables the formation of shell shapes from sheet materials, and used carbon fiber reinforced thermo plastic (CFRTP) for the samples. In the study, a laser-assist incremental forming system based on 3D computer-aided design (CAD) data was developed. The system comprises computer-aided manufacturing (CAM) system, which generates a tool path based on CAD data and evaluates the formability between the CAD data and 3D-scanned data, including alignment compensation. The feasibility of the developed system was demonstrated through a set of forming experiments.

Efficient hybrid joining methods are required for joining metals and plastics in the automobile and airplane industries. Injection molded direct joining (IMDJ) is a direct joining technique that uses metal pretreatment and injection molding of plastic to form joints without using any additional parts. This joining technique has attracted attention from industries for its advantages of high efficiency and low cost in mass production. Blast-assisted IMDJ, an IMDJ technique that employs blasting as the metal pretreatment, has become suitable for the industry because metal pretreatment can be performed during the formation of the metal surface structure without chemicals. To satisfy industry standards, the blast-assisted IMDJ technique needs to be optimized under blasting conditions to improve joining performance. The number of parameters and their interactions make this problem difficult to solve using conventional control variable methods. We propose applying statistical and artificial intelligence analyses to address this problem. We used multiple linear regression and back propagation neural networks to analyze the experimental data. The results elucidated the relationship between the blasting conditions and joining strength. According to the machine learning predicted results, the best joining strength in blast-assisted IMDJ reached 22.3 MPa under optimized blasting conditions. This study provides new insights into similar engineering problems.

This study proposes a method for automating the determination of assembly order by automating the derivation of the necessary connection relationships between the parts. The proposed method minimizes the information required for the initial conditions and automatically determines the feasible assembly orders. As a general rule, based on the assumption that the assembly order for a product is the reverse of the disassembly order, once the disassembly order is derived based on the 3D CAD model and the connection relationships between the parts, the assembly order can be determined. Until now, however, the relationships between the parts are decided manually by the attendant engineers, thus, hindering the full automation of the determination of the assembly order. To achieve full automation realistically, the connection relationships between the parts should be derived automatically from the 3D CAD model, for which this study proposes an efficient method. The components were extracted from the 3D CAD model, and the bolts were identified. The connection relationships between the parts were derived from the interference conditions determined while moving each part minutely. An association chart diagram was created from the obtained connection relationships, from which multiple assembly order candidates could be derived.

Transforming assemblies are products that alter their shapes by re-assembling their parts. This idea is applied to a wide range of objects from folding gadgets as outdoor gear aimed at saving space, to robotic characters fighting in Hollywood films which drastically change their appearance. While the former type falls into a folding or packing problems, the latter requires a different viewpoint to be solved since the destination shape is not necessarily aiming at minimizing the occupation space. As a possible solution, this kind of deformation can be decomposed into segmentation of the shape to parts and parts matching. Segmentation is a general problem in shape modeling and numerous algorithms have been proposed for this. On the other hand, matching simultaneously multiple parts (many-to-many matching) has hardly been explored. This study develops a many-to-many matching algorithm for surface meshes of parts from two distinct destination shapes of a single transforming assembly. The proposed algorithm consists of a local geometry analysis and a global optimization of parts combination based on such analysis. For the local geometry analysis, the surface geometric feature is described by a local shape descriptor. Some vertices are detected as feature points by intrinsic shape signature (ISS) and the geometry at the feature points is expressed by the signature of histogram of orientation (SHOT). For all the combination of pairs from each destination shape, the number of feature points with similar descriptor values is counted. In the global optimization, the final matching is determined by the maximum weight matching on a complete bipartite graph whose nodes are the parts, and edges are weighted by the number of the feature points with similar descriptors. We present successful results for several examples to empirically show the effectiveness of the proposed algorithm.

The Future Design (FD) workshop (FDWS) is a discussion framework based on FD. The aim of FD is to activate a human trait called futurability, considering the preferences of future generations. Previous FD practices with the theme of policy-making in local governments have demonstrated this possibility. However, creating concrete proposals might depend on workshop participants’ abilities and emotions to perceive future society. By comparing two case studies, this study examines the effects of a method for utilizing a causal loop diagram (CLD), a tool for systems thinking, in FDWS to systematically draw the future society and activate discussions among the participants. CLD is a qualitative system model that helps identify the factors that lead to systemic problems and analyze the guidelines for solving them. Its effects on the performance of the FDWS discussion activity are evaluated. They are quantified by text mining analysis using participants’ remark records. Two case studies conducted at policy-making workshops in the local governments of Japan are examined. One is the FDWS in Kyoto City which adopted the proposed CLD utilization method, and the other is the FDWS in Suita City without CLD. The comparative analysis demonstrates that the proposed method makes the discussion livelier, less divergent, and more developed in the FDWS.