Health Monitoring of Electro-Pneumatic Controlled Systems Using Multivariate Latent Methods: An Experimental Validation

Electro-Pneumatic systems exhibit highly nonlinear characteristics due to air compressibility, the presence of friction and the nonlinearities of control valves. Monitoring by acquiring the system's transfer function accurately can be difficult for nonlinear systems. This paper outlines a new idea that one can deal with the electro-pneumatic system as a black box, and using a multivariate technique called principal component analysis (PCA) and projection to latent structure discriminant analysis (PLSDA) to provide robust information about the system's condition. The monitoring system has been experimentally validated for an electro-pneumatic printing machine system using vibration, pressure and displacement sensory data integration using PCA-PLSDA algorithm. Experiments were conducted under two pressures for three artificially conditions: normal, throttled, and leaking system. The models were tested under these conditions, and the results showed that the proposed technique can successfully differentiate between these process conditions and can be used to overcome these faulty states using external connection line in a real-time environment.