This paper surveys the artificial intelligence methodologies which can be applied to electromagnetics and mechanics. This paper describes what AI and machine learning can do for engineers in these fields. It also presents the principles, recent advances and future outlooks of AI applications. As examples, anomaly detection, dentification of material properties and machine characteristics are discussed. Moreover, key points to make effective use of AI for research and development of electric and mechanical systems are described.

　This paper describes a motor characteristic prediction method using deep learning and accelerating topology optimization. Convolutional neural networks, one of the deep learning methods, have been shown to be able to accurately predict motor characteristics for shapes that are handled by topology optimization. In addition, the applicability of explainable deep learning to the design of motors is discussed.

　An automatic design method of electric machines has been developed based on tree structure and reinforcement learning. The decision process of electric machines designs is modeled as the tree structure, and values of the nodes are learned from the design data generated by expert designers. The candidate designs are automatically generated through the tree search based on the node values. The designers’ knowledge and experience can be utilized in this automatic design method through the learning of the node values.

　This paper presents a novel total design method using Monte Carlo tree search and hybridized topology-parameter optimization. In this method, the global structure of an electrical machine is represented by nodes in a tree structure. Monte Carlo tree search is used to select a path in the tree structure, which corresponds to a machine structure. After that, detailed machine shape is optimized by the hybridized method considering both of flexibility and manufacturability. By repeating the above procedure, the global structure and detailed shape are simultaneously optimized. This method is applied to the design of a permanent magnet motor to enhance torque characteristics.

　Recently, many studies using deep learning have been reported in the field of design optimization of electric machinery based on electromagnetic field analysis. It is expected to bring unprecedented sophistication to design technology by performing complementary combination of the deductive finite element analysis with an inductive approach that constructs a Blackbox model between input and output through learning based on a large amount of data. This paper introduces some examples of effective use of both in the field of electric equipment design.

　This paper presents an approach to extracting visual features from practical rotating machines based on a convolutional neural network (CNN). We first generate synchronous reluctance motor images via topology optimization using two objectives: maximizing average torques and minimizing torque ripples. Each image is assigned two class labels based on its average-torque and torque-ripple values. Then, using the pairs of images and their two types of labels, we train a CNN based on multi-task learning that simultaneously predicts the two types of classes. Finally, we visualize the features learned by the CNN using a class activation mapping method.

　This paper presents novel multi-material topology optimization methods for hybrid-type motors. To improve the hybrid-type motors, two type optimization methods are proposed. In the first optimization method, optimization process is divided into two steps, and different material distribution is optimized on each step. In the second optimization method, combinatorial optimization is used to aid in the solution search. To validate the effectiveness, the proposed methos are applied to the hybrid-type motors.

　Transcutaneous energy transfer system (TETS) for ventricular assist device (VAD) is a safe and effective technique that allows power to be delivered to implants wirelessly, avoiding the use of percutaneous wires. TETS is sensitive to mutual electromagnetic interference when multiple TETSs concentrate in the same location which may danger to device wearers. In this paper, two TETSs will be arranged in eight different configuration patterns to verify if it is possibility that mutual electromagnetic interference can enhance biological effects of human body by measured the internal electric fields (E_INT) and TETS output voltages (V_out). The operating frequency of TETSs were fixed to 400 kHz. A novel four-layer human body simulation models consist of dry skin, wet skin, fat, and muscle were used as the TETS’s wearers. The results indicated that E_INT was increased by 16.4% and V_out was increased by 15.5% in the worst case. To ensure that TETS can be used safely, the distance between wearers should be greater than 20 cm within the simulations.