建築熱源システムを対象としたデータ駆動型不具合検知・診断　（第4報）畳み込みニューラルネットワークによる不具合の程度と複数不具合の診断

抄録

We examined fault detection and diagnosis (FDD) by convolutional neural network (CNN) using the training database of faults generated in the previous report. CNN which learned all cases could diagnose accurately and those learned only single faults could diagnose a fault which has large influence on the system.