Projects / Programmes
Combined Qualitative/Quantitative Approach to On-Line Fault Diagnosis of Industrial Processes
Code |
Science |
Field |
Subfield |
2.06.01 |
Engineering sciences and technologies |
Systems and cybernetics |
Control systems technology |
Code |
Science |
Field |
T120 |
Technological sciences |
Systems engineering, computer technology |
T125 |
Technological sciences |
Automation, robotics, control engineering |
fault diagnosis, fault detection, fault isolation, model-based methods, approximate reasoning, parity relations, qualitative modelling, fault-trees, toolboxes, prototyping, benchmarks, process monitoring, industrial applications.
Researchers (9)
no. |
Code |
Name and surname |
Research area |
Role |
Period |
No. of publicationsNo. of publications |
1. |
15582 |
Janez Grom |
|
Researcher |
1997 - 2000 |
14 |
2. |
02561 |
PhD Đani Juričić |
Systems and cybernetics |
Head |
1997 - 2000 |
416 |
3. |
16160 |
Mateja Kavčič |
Systems and cybernetics |
Researcher |
1997 - 2000 |
25 |
4. |
16190 |
PhD Andrej Rakar |
Metrology |
Researcher |
1997 - 2000 |
67 |
5. |
05191 |
PhD Marjan Rihar |
Engineering sciences and technologies |
Researcher |
1998 - 2000 |
96 |
6. |
02830 |
PhD Stanislav Strmčnik |
Systems and cybernetics |
Researcher |
1997 - 2000 |
488 |
7. |
15583 |
Miroslav Štrubelj |
|
Researcher |
1999 - 2000 |
30 |
8. |
14018 |
PhD Mina Žele |
Computer science and informatics |
Researcher |
1997 - 2000 |
56 |
9. |
12343 |
PhD Alenka Žnidaršič |
Systems and cybernetics |
Researcher |
1997 - 2000 |
80 |
Organisations (1)
no. |
Code |
Research organisation |
City |
Registration number |
No. of publicationsNo. of publications |
1. |
0106 |
Jožef Stefan Institute |
Ljubljana |
5051606000 |
91,767 |
Abstract
Detection and localisation of sources of malfunctioning is one of the most important tasks in process monitoring. Current generation of SCADA systems provide merely classical alarming based on violation of upper and lower bounds imposed on measurable process variables. These bounds are usually set up in a heuristic manner. Thus mainly large faults can be detected, unfortunately, often too late to take corrective actions. Therefore, in order to early reveal a fault in the system, it is neccessary to make use of more profound knowledge of the plant. Depending on the diagnostic requirements, the adequate purposive models might take various levels of abstraction.
The purpose of the underlying project is the following:
1. design and combination of diagnostic algorithms based on models of various levels of abstraction,
2. robust analytical model-based diagnosis employing the quantification of modelling error,
3. contribution to the computer-aided design for prototyping of diagnostic systems and
4. practical realisation of diagnostic systems for various test and industrial processes.
The most relevant results achieved could be summarised as follows:
- application of stochastic embedding technique in order to quantify the contribution of modelling errors to the variance of the prediction error; consequently, a likelihood ratio test for detection is derived,
- component-based approach to process modelling utilising the naive physics approach,
- fault isolation by means of approximate reasoning with focus on the transferable belief model (TBM),
- contribution to the toolbox in G2 for the synthesis of diagnostic rules and
- testing of various diagnostic solutions on real processes including the diagnostic prototype for a subprocess in TiO2 production plant in Cinkarna Celje.