Main contents

Control Engineering:
Operations & Maintenance

Performance monitoring and fault diagnosis of industrial processes and machines

Monitoring the performance of an industrial process or of a machine is usually of great interest to the owner, since maximizing performance usually maximizes also quality and income. Different types of performance measures can be constructed. Their basic use is in monitoring, but this can be extended to condition monitoring and fault diagnosis in order to find the causes of poor performance as well as components that need maintenance.

Fault detection and condition monitoring methods are usually divided into three main categories: model-based, knowledge-based and data-driven systems. In model-based methods a mathematical model (often a differential equation system) is built for the monitored system from physical principles and the predictions given by this model are compared with the actual measurements from the system to detect faults. The main limitation with this approach is that many systems are too complex to allow any sufficiently accurate model to be built. Often fault situations are handled by ad-hoc procedures developed by human experts. Knowledge-based methods (e.g. expert systems, fuzzy logic) try to automate the use of this knowledge. The problem with these methods is that it is laborious to gather the knowledge from the experts and to maintain the database as products evolve. The approach used in this work is data-driven. Data-driven methods try to deduce the properties of the system more or less directly from the available measurement data, which are usually readily available in modern industrial plants.

Schematic representation of process monitoring

The schematic above represents a complete process monitoring loop. During normal operation a fault detection system monitors the process for possible faults. If a fault is detected, the affected measurement signals are isolated in fault identification. Then the exact location, magnitude, and time of the fault are found out during the fault diagnosis phase. The corrective actions that return the system to its normal operating state are then performed (system recovery).

Active projects

  • MINMO — Advanced Methods in Monitoring and Control of Ore Concentration
  • OTTO — Optimizing the work-efficiency of machine operators
  • WiSA — Wireless Sensor and Actuator Networks for Measurement and Control

Past projects

  • AKKU — Optimal Supervision And Use of Battery Systems
  • AULA — Quality and competitive edge with innovative automation
    • LATU — Improving quality and productivity in forest machine work and timber harvesting chain
    • PALASET — Support system for intelligent monitoring of quality variation in paper making
    • RIKE — Performance Improvement for Concentration Process
    • VIATON — Advanced Modelling and Control of the ViaFill Process
  • ERHE — Very Distributed Sensor and Actuator Networks
  • ICT-E — ICT of Electric distribution network
  • SOSE — Soft-sensor methods in improving the competitivity of industrial products
    • Harvester — Intelligent maintenance and performance optimization of forest machines with soft-sensor methods
    • VÄSY — Intelligent, Machine Vision Based Control for a Flotation Process
    • VAIVI — Fault diagnostics of electrical AC machines
    • PIHA — Circuit manufacturing control
  • TILLIKKA — Improving industrial competitiveness with condition monitoring
    • NYKY — Application of modern condition monitoring and control concepts in forest machines
    • ÄKSY — Intelligent methods in mining environment
    • MESTA — Model Based Estimation Methods in Analysis of Electronics Manufacturing
    • LAMA — Computational methods in the monitoring of satellite AOCS
    • DADA — Development of data fusion and diagnostics methods in weather station networks
    • PALA — Advanced monitoring of paper quality