Fault Detection of Actuator with Digital Positioner Based on Trend Analysis Method

Abstract

The demanding on high safety, performance and reliability in controlled processes has becoming increasingly stringent in recent years. Control valves (or actuators) are widely used in industrial processes. As the final control elements, they are often installed in the technology nodes working in the harsh environment: high temperature, high pressure, humidity, pollution, chemical solvents, etc. Their malfunctions usually lead to poor control performance or process disturbance, even result in unqualified product. Therefore, the online detection and diagnosis of control valve should be applied to preserve the highreliability of control valves due to the severity of its possible effects of failure on the processes. The malfunctions of actuator mainly include fully failure, offset and bias, change of gain, serious hysteresis, and stick-slip fault. In the past two decades, there has existed a number of fault detection and diagnosis methods for actuators in process control systems. Some efforts involved model-based approaches: state estimation (Hoefling et al.,1995; Park T.G. et al., 2000; Edwin Engin Yaz & Asad Azemi,1998); parity equation (Massoumia et al.,1998; Mediavilla et al.,1997). These methods require relatively accurate mathematic models about the processes. However, it is very difficult to obtain accurate mathematic models in most industrial processes. Other studies focused on using neural networks (Patan,2001; Patan & Parisini,2003; Pawel et al.,2003), fuzzy logic, and signal analysis (Deibert,1994). An important issue that should be highlighted is that, there is no a method that can detect and diagnose all kinds of faults because various fault types may occur in control valves. most existing method requires process knowledge or user-interaction (Forsman & Stattin 1999; Hagglund,1995; Wallen,1997). Only few approaches do not need prior knowledge about the process (Horch, 1999). In process control systems, actuators with digital positioner are widely used. Generally, few related signals can be sampled to process monitor systems. These signals are: the input and output signals related to the component itself, and the flow signal that controlled by the industrial actuator. In fact, these signals provide useful information about the operation of the actuator. In this chapter, a series of methods based on trend analysis are proposed to detect typical faults of industrial actuators with digital positioners by using these three signals. Because 14