Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Particle filtering on hybrid dynamical systems for sensor fault detection
Thesis (MEng)--Stellenbosch University, 2024.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613773690306560 |
|---|---|
| access_status_str | Open Access |
| author | Loubser, Simone Diana |
| author2 | Louw, Tobias |
| author_browse | Loubser, Simone Diana Louw, Tobias |
| author_facet | Louw, Tobias Loubser, Simone Diana |
| author_sort | Loubser, Simone Diana |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description |
Thesis (MEng)--Stellenbosch University, 2024. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/131808 |
| institution | Stellenbosch University (South Africa) |
| last_indexed | 2026-06-10T12:41:28.315Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/131808 Particle filtering on hybrid dynamical systems for sensor fault detection Loubser, Simone Diana Louw, Tobias Bradshaw, Steven Stellenbosch University. Faculty of Engineering. Dept. of Chemical Engineering. Detectors -- Calibration Hybrid systems Discrete element method UCTD Thesis (MEng)--Stellenbosch University, 2024. ENGLISH ABSTRACT: Process variables are commonly assumed to be continuous random variables in models of chemical process plants. Sensors that provide these measurements are susceptible to faults. These sensor faults are common in the chemical engineering field and can distort the observation of the process. Monitoring sensor health, and potential sensor faults, requires monitoring of the state of the sensor. These sensor states may be modelled as discrete random variables jointly distributed with process states and measurements, resulting in a hybrid (continuous and discrete) dynamical system. There is a growing interest in process monitoring solutions that provide rapid and reliable fault detection and diagnosis (FDD), for a wide range of abnormalities. State estimation of hybrid dynamical systems is challenging due to the combinatorial explosions following propagation of the discrete states. For this reason, a new algorithm has been developed that utilizes the structure of the particle filter to incorporate discrete states representing sensor health: Particle Filter with Fault Detection and Diagnosis (PF-FDD). This method is tailored to the FDD of sensor faults and overcomes the challenges of combinatorial explosions in discrete state propagation. This approach uses the number of sampled particles to succinctly represent the prior distribution of sensor states, and then tracks the sensor transitions whilst simultaneously tracking the process states, for each particle. The PF-FDD method detects and diagnoses the highly specific abnormal process behaviour caused by a sensor fault, with high sensitivity. The discrete sensor states are inherently categorical, representing either normal or faulty conditions, with distinct faults (e.g., stuck, biased, etc.) defined as individual states. The method has been evaluated using a simulated continuous stirred tank reactor (CSTR) benchmark with simulated datasets spanning 15 and 100 days. These simulations varied in fault models, process complexity and disturbances, all of which were used to test the robustness of the PF-FDD approach. The PF-FDD method showed high accuracy in state estimation, with a maximum absolute estimation error (MAPE) of 15% across all the simulations, and lowest of 0.15%. Furthermore, it was found that the PF-FDD method could successfully provide promising FDD results. It detected and correctly diagnosed sensor faults, with high sensitivity. The detection sensitivity consistently exceeded 60% and specificity was above 90%, in all applications. The approach excelled at diagnosing stuck and failed sensor fault types, with more than 70% and 60% correct diagnoses in majority of the simulations, respectively. However, the biased sensor fault type had very low detectability and thus was diagnosed with low sensitivity: less than 60% in all simulations. This method relies on an accurate system and fault models to effectively estimate the state of the system and diagnose faults. To overcome uncertainties in the fault models, the number of particles could increase, but this will impact the computational expense of the method, making FDD of complex systems challenging. Overall, the PF-FDD approach provided reliable and robust performance, especially for well-defined fault models, in the presence of varying fault and process complexities, as well as during simultaneous fault occurrences. AFRIKAANSE OPSOMMING: Geen opsomming beskikbaar. Masters 2025-03-28T09:46:30Z 2025-03-28T09:46:30Z 2024-12 Thesis https://scholar.sun.ac.za/handle/10019.1/131808 Stellenbosch University xiv, 137 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Detectors -- Calibration Hybrid systems Discrete element method UCTD Loubser, Simone Diana Particle filtering on hybrid dynamical systems for sensor fault detection |
| title | Particle filtering on hybrid dynamical systems for sensor fault detection |
| title_full | Particle filtering on hybrid dynamical systems for sensor fault detection |
| title_fullStr | Particle filtering on hybrid dynamical systems for sensor fault detection |
| title_full_unstemmed | Particle filtering on hybrid dynamical systems for sensor fault detection |
| title_short | Particle filtering on hybrid dynamical systems for sensor fault detection |
| title_sort | particle filtering on hybrid dynamical systems for sensor fault detection |
| topic | Detectors -- Calibration Hybrid systems Discrete element method UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/131808 |
| work_keys_str_mv | AT loubsersimonediana particlefilteringonhybriddynamicalsystemsforsensorfaultdetection |