Monitoring diagnostic apparatus using neural network

A monitoring diagnostic apparatus for detecting abnormalities occurring in a monitored object comprising:

• an acceleration sensor installed at said monitored object for measuring an acceleration of the monitored object resulting from vibration occurring in said monitored object and outputting an acceleration signal representing said acceleration;
• a signal processing means for executing predetermined signal processing on said acceleration signal;
• a neural network for identifying an abnormality and determining an abnormality cause thereof by inputting a signal undergoing said signal processing at said signal processing means:
• wherein said neural network comprises a plurality of neural elements constituting an input layer, a plurality of neural elements constituting an output layer and connectivity weights connecting said input-layer neural elements to said output-layer neural elements, and wherein said input-layer neural elements are used for inputting signals generated by said signal processing means;
• wherein the values of said connectivity weights of said neural network are determined in a learning process so that a specific output-layer neural element produces a maximum output corresponding to an already known cause of an abnormality occurring in said monitored object;
• and
• wherein in the event of an abnormality occurring in said monitored object whose cause is not known in advance, said neural network undergoes a learning process for self-creating a new output-layer neural element and determining the values of said connectivity weights so that said new output-layer neural element produces a maximum output.