The RT method, one of the methods of the Mahalanobis-Taguchi (MT) system, has been applied to various fields. The RT-PC method is another RT method that incorporate principal component analysis. This method can be applied to data set consisting of variables that do not have the same dimensions.

This study proposes an RT-PC method that incorporate sparse principal component analysis (SPCA) as an anomaly detection method for small-sample data. SPCA incorporates L1 regularization term into traditional principal component analysis, where an appropriate choice of L1 regularization term improves the accuracy of anomaly detection for sparse data and reduces the number of nonzero elements in the eigenvectors, thereby increasing interpretability. Numerical experiments have confirmed that the accuracy of the proposed procedure is similar to or better than conventional procedure, and we have confirmed that an RT-PC method that incorporate SPCA reduces the number of nonzero elements in the eigenvectors more than RT-PC method. This is thought to improve interpretability of principal component.

Obtaining the level of a factor that has the greatest effect on product development or crop cultivation is important given limited time, money, labor, and resources. Design of experiments is an approach to efficiently collect data in order to clarify the relationship between factors and effect. However, design of experiments mainly ends with the initial experimental design, which makes it difficult to conduct several more experiments based on the results. On the other hand, Bayesian Optimization, one of the methods of inverse estimation, uses Latin Hypercube design as the initial design of experiments and then conduct experiments sequentially. This is a powerful advantage of Bayesian Optimization.

Bayesian Optimization involves several factors including the kernel function, the acquisition function which evaluates experimental points, the initial design of experiments, and the total number of experiments. We clarify how many initial experimental points should be taken and whether the number of initial experimental points depends on the characteristics of the benchmark functions.

This research examines whether it is better to choose the optimal solution in Bayesian Optimization from the experimental points or from the mean value of the learned Gaussian Process Regression model. We provide guidance for users of Bayesian Optimization.

The design of machining fixtures affects the lead time for product development. Therefore, it is necessary to shorten the design time while maintaining the required quality. However, it is difficult to identify bottleneck processes and implement countermeasures because estimating the individual work hours for each process is challenging. This is attributed to the characteristics of the design work, such as multitasking and repetitive activities. In this study, person-hours were identified as a critical factor. Through interviews with designers, we clarified the bottleneck processes and developed a method for shortening the design time. Subsequently, we applied an actual design to verify its effectiveness. The method involves detailing the machining fixture design for a specific design target and designing and managing individual plans, including the identification of bottleneck processes and the implementation of countermeasures. The proposed method was applied to a machining fixture design consisting of seven parts and twelve processes. The predicted value reduced the design time by 53%, whereas the actual measured value reduced the design time by 59%. This confirmed the validity of the proposed method.