保全分野におけるAIを用いた順解析、逆解析、未来予測の試み
概要
Abstract As an attempt to apply AI in the field of maintenance, we have been studying "forward analysis", "inverse analysis" and "future prediction". As for the “forward analysis”, as a theoretical verification of the digital hammering inspection results using the AE sensor, the natural frequency obtained in the hammering inspection is confirmed by “time history response analysis”. Because there is a combination of the shape, material properties to be inspected, measuring position of natural frequency, etc., it is necessary to perform enormous amount of analysis. Therefore, learning the relationship between typical analysis conditions and analysis results using AI eliminates the need to perform the analysis each time. As for “inverse analysis”, we report the case of evaluating the restraint state of the inspection object from the digital hammering inspection results and the shape of the object. By learning the relationship between “natural frequency”, “shape of the inspection object” and “restraint state of the object” by AI, the restraint state can be calculated immediately from the natural frequency and the shape. With regard to "future prediction", changes in the "objective function" when changing the "design variable" are predicted based on the relationship between the "design variable" and the "objective function" learned by AI in advance.
