Efficient Industrial Process Operation Through Fault Diagnosis

Maintaining a competitive edge in the manufacturing sector requires process performance monitoring and abnormal situation management. Despite the increasingly common use of advanced control systems (e.g. model predictive control tools), some sectors, notably the pulp and paper industry, continue to face abnormal situation management issues. This challenge is due to the fact that the few existing commercial tools are primarily intended for expert users and therefore do not help control room operators in understanding the abnormal situations or in diagnosing anomalies in order to restore the normal functioning of a process.

Automatic Fault Detection and Diagnosis System

CanmetENERGY experts are working toward developing and implementing an automatic fault and detection diagnosis system that will:

  • Perform an online analysis of operating data
  • Provide users with useful and structured information
  • Allow decision making

This system will use a (qualitative and quantitative) hybrid method and will systematically integrate innovative software tools to allow operators to better understand the process and its control conditions, so as to improve the overall plant functioning and management of abnormal situations. By identifying possible malfunctions early on in the process, operators will be able to take immediate corrective action, thereby ensuring that the process is operating within normal limits, and also to improve the equipment maintenance plan, which will in turn reduce energy consumption and increase productivity.

The system will consist of generic and modular computerized toolboxes that will be easily reusable, at low cost, in several other industrial applications. It will also be capable of automatic learning, so as to continuously feed new information into the database in order to improve system capacity. The system will include the following toolboxes:

  • A knowledge base module with qualitative and quantitative data modelling
  • An inference engine, including malfunction detection and classification agents
  • A diagnostic procedure using various agents to identify malfunctions
  • A user interface with the required communication capacities

A method to determine how the various toolboxes may be used either individually or together will also be developed, allowing new applications for different processes to be easily built.

CanmetENERGY is seeking partners interested in collaborating on this research project. To discuss a potential partnership, please contact us.

Process detection & diagnosis system (in blue) to supervise existing control system (in orange).

Process detection & diagnosis system (in blue) to supervise existing control system (in red).

Larger image

Text Version

Process detection & diagnosis system

The following image represents a process detection and diagnosis system overseeing an existing control system. The various steps are connected by arrows indicating the order of operations and the links between them.

On the left, there is a diagram of an existing control system with its three elements (the Existing Process Control, the Process, and the Data) all linked by arrows. From the Data element, an arrow points to a process detection and diagnosis system represented on the right. The first component of this system is that of Data Mining; secondly, there is the Process Monitoring and Fault Detection step, followed by the Fault Diagnosis stage of the operation; and finally, there is the Decision Support step. Following this last step, improvements are either automatically implemented in the process or made by a system operator who physically administers the suggested changes.