Napredna pretraga

Pregled bibliografske jedinice broj: 41542

Fault detection of industrial processes

Valh, Drago
Fault detection of industrial processes 2000., magistarski rad, School of Environmental Sciences, Nova Gorica

Fault detection of industrial processes

Valh, Drago

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, magistarski rad

School of Environmental Sciences

Nova Gorica




Tovornik, Boris ; Vukić, Zoran

Ključne riječi
Process safety; occupational safety; safety analysis; instrument fault detection; component fault detection; heat exchanger; state observers; parameter estimation; fuzzy reasoning; alarm analysis; SCADA systems

The work is divided into two parts. The first part deals with a safety of (dynamical) process industry in general. The three main hazardous events with their causes and consequences are shortly described. The section points out the idea that safety analysis should predict parts of the process where early fault detection is indespensable. This is important to have proper sensors installed. Early fault detection can improve the quality of production and increase the occupational and environmental safety through detecting the faults before they develop to a failure that might cause an accident with all its consequences such as material detriment, environmental damage or loss of human life. Second part briefly describes some model-based fault detection techniques and an interesting approach - alarm analysis. Some of the methods were tested on the pilot plant of the laboratory model of the heat exchanging device. An analitically derived dynamic model and an experimentally obtained static model were used to test the parity space approach for residual generation. Sensor faults and two component faults (valve(tube) obturation was detected using parity space approach, and the leakage of the vessel was detected using non-linear estimators) were detected and isolated succesfully. The idea of decoupling the process into smaller (analytical) parts and to use as many measured signals as possible as inputs to partial models to overcome the modelling uncertainties as much as possible was implemented. The propagation of a fault is seen from the analytical analysis of the process (mathematical model) and the knowledge is used in the fuzzy reasoning membership table. A heuristic approach called alarm analysis is suggested as an alternative approach to analytical redundancy.

Izvorni jezik

Znanstvena područja


Projekt / tema

Fakultet elektrotehnike i računarstva, Zagreb