In process industry, the demand for higher product quality, better process efficiency and an overall increased system performance, transforms the engineering problem of fault detection and diagnosis into a high priority task. A single untreated fault may have serious consequences to a process or a whole industry; the United States petrochemical industry reports that faults cause a reduction of 3% to 8% of oil production yearly, the economic impact being approximately 20 billion of dollar losses.
The actions of detecting an abnormal event in the process, diagnosing its root causes and taking the correct control action to bring the process back to a normal operating state have fallen, until recently, in the hands of the operators. However, as processes grow in size and complexity it is difficult for a human operator to take the necessary corrective actions in a timely manner. In this context, the developed automated diagnosis methods have become essential to keep complex modern processes at a high level of efficiency.
Fault diagnosis objective is the determination of a fault giving as much information as possible about the characteristics of such fault. Detecting, isolating and identifying the cause of a fault are common actions that any fault diagnosis systems must perform. The information required to perform these actions and the manner in which the diagnosis information is presented to the final user are the main differences between fault diagnosis systems.
Nowadays in literature there is plenty of information about fault diagnosis systems, and the different faults present in the process industry. Also different proposed classifications for faults and different classifications for fault diagnosis methods are proposed. Nevertheless there is no classification related to the types of faults, with the information needed for a fault diagnosis system to perform their task efficiently.
The purpose of this research is to propose a classification of the faults found in process industries in relation to the fault diagnosis methods that have been proven the most effective to diagnose them.