Investigations on diagnosis and prognosis of faults in pneumatic actuator using machine learning techniques

Abstract

Process industries involve at least one control loop which hosts in newlineitself many devices and equipment for accomplishing various tasks. About newline40% of the fatal accidents in these industries are due to the faults and failures newlineof the critical equipment installed in the various process control loops. The newlinemost accident-prone equipment is the process equipment like sensors, newlineactuators, transmitters etc. These critical process systems should comply with newlinethe safety standards to avoid any abnormal event progression and productivity newlineloss. Thus, industries expect safety and transparency as the two functionalities newlineduring their operation. With this objective in mind, a systematic approach to newlinediagnose faults and predict the lifetime of one among the critical equipment newlinenamely, the pneumatic actuator is proposed. newlineDiagnosis of faults in pneumatic actuators is vital for the effective newlinestudy of the actuator fault profiles, detection and identification. Therefore, a newlineclear idea of the type of faults, occurring instances, magnitude, intensity and newlineplace of fault is required. This allows the operator to eliminate the faults at the newlineearliest. Fault diagnosis methods are classified into model-based fault newlinediagnosis, data driven fault diagnosis and hybrid fault diagnosis. newlinePrognostics is vital to detect the faults and the failure rate and newlineestimate the remaining useful life period of the equipment. Remaining Useful newlineLife (RUL) of equipment is defined as that period when the equipment newlineestablishes itself in its fullest working condition after sometime of usage. newline