A practical algorithm for diagnosing control loop problems

This paper describes an algorithm for detecting and diagnosing poor performance in feedback control loops. The performance of the control loop is assessed based on criteria derived from parameters that can be estimated from superficial knowledge of the controlled process. The algorithm can distinguish between three types of control problems: oscillatory response, sluggish response, and plant failure. The method is particularly applicable to heating, ventilating, and air-conditioning (HVAC) systems where linear (PI) control law is normally used despite the existence of often quite non-linear operating characteristics. In these systems, the control performance varies as the plant is exercised across its operating range. In practice, periodical re-tuning of HVAC control loops during the yearly load cycle may be necessary in order to obtain consistent control performance. The algorithm presented in this paper can act as the basis for determining when re-tuning is necessary in addition to detecting plant failure. Potential advantages of detecting control problems in HVAC systems are energy savings, reduced maintenance costs, and improved thermal comfort conditions. The paper describes the techniques comprising the algorithm and presents results from tests carried out on a simulated air-handling unit (AHU).