In machining, errors occur due to the complex effects of multiple factors, such as thermal deformation of tools and workpieces, deformation due to cutting forces, and deterioration of machine tools. To reduce machining errors, it is therefore necessary to consider the interaction of these multiple factors. In recent years, studies have been conducted on the prediction of machining errors by combining machine learning with monitoring data that can be obtained through digitization. Most of these studies, however, only predict machining errors and do not analyze the error factors. In this study, we developed a system that enables factor analysis by identifying the variables that have the highest accuracy in predicting machining errors in machine learning models. We then visualize the reasons behind error prediction using an interpretation method for the machine learning models created from the variables. We demonstrate the effectiveness of the proposed system in an experiment using an end-milling machine.

The corner module, which integrates an in-wheel motor and a steering motor, is attracting attention as a means of building a powertrain for electric vehicles. The corner module enables independent control of the braking/driving and steering forces of each wheel, which provides a high degree of freedom and redundancy, and is expected to dramatically improve vehicle dynamics and fault-tolerance performance through all-wheel drive and all-wheel steering. On the other hand, even if redundancy is established mechanically, it is meaningless as a redundant system unless the redundancy is well used. In other words, there is an urgent need to establish a control method that allows corner modules to operate appropriately according to driving conditions. In this study, we apply broadcast control, one of the multi-agent control theories, to construct a four-wheel independent steering control system. Through numerical simulations, we confirm that the proposed method brings out the effect of the redundant system and improves the vehicle performance.