Abstract
Distributed parameter systems (DPSs) widely exist in industrial processes (e.g., transport-reaction processes, thermal processes, and fluid processes). One unique feature of DPSs is the so-called spatio-temporal dynamics, which demonstrates that the system input, output, and even parameters can vary both in the space and time domain. With the fast-growing demands for high production quality as well as economic operations on complex industrial manufacturing systems, the requirements of system safety and reliability become more and more critical. Hence, abnormality detection and localization of complex industrial manufacturing processes described by DPSs is extremely important. This thesis is mainly concerned with the abnormality detection and localization of DPSs in both model-based (Part I) and data-driven (Part II) perspectives:Part I: Model-Based Perspectives:
1. Spectral-Approximation-Based Abnormality Detection and Localization for Semi-linear Parabolic DPSs
Considering the spatial distribution characteristic of distributed parameter systems (DPSs), the abnormality localization problem is physically intuitive but seldom studied. We make the first attempt to systematically study the abnormality localization problem for a class of semi-linear parabolic DPSs without knowing prior information of spatial distribution function of the spatio-temporal (S-T) abnormality. An abnormality detection and localization filter is first designed to generate a distributed residual signal using limited sensor measurements with the help of the spectral approximation. The residual evaluation procedure is also conducted in a distributed manner.
2. Backstepping-Based Distributed Abnormality Localization for Linear Parabolic DPSs
To deal with the observation spillover introduced by the spectral approximation, a systematic framework is proposed to address the distributed abnormality localization problem for a class of linear parabolic DPSs using limited in-domain measurements plus one boundary measurement rather than the full state measurement. The proposed methodology consists of an abnormality detection filter (ADF) and an abnormality localization filter (ALF) design based on the backstepping techniques. For the detection purpose, the residual is evaluated in a lumped manner; For the localization purpose, a distributed residual is first constructed and is evaluated in a distributed manner.
3. Adaptive Observer-Based Abnormal Source Identification for Linear Parabolic DPSs
Identification of abnormal source hidden in DPSs belongs to the category of inverse source problems. It is important in industrial applications but seldom studied. We make the first attempt to investigate the abnormal spatio-temporal (S-T) source identification for a class of DPSs. An inverse S-T model for abnormal source identification is developed for the first time. It consists of an adaptive state observer for source identification and an adaptive source estimation algorithm. One major advantage of the proposed inverse S-T model is that only the system output is utilized, without any state measurement.
Part II: Data-Driven Perspectives:
4. Dynamic-Spatial-Independent-Component-Analysis-Based Abnormality Localization for DPSs
A novel data-driven approach is proposed to localize abnormality for DPSs in a data-driven manner. The cross-correlation order of DPSs in the space domain is firstly obtained by the cumulants-based identification method. Then a spatial augmented matrix of the spatial-temporal distribution data is formed and dynamic spatial independent component analysis (dynamic SICA) method is proposed for independent decomposition. The dominant spatial independent components will be extracted and the spatial residuals can be produced for spatial reference statistics. Through the kernel density estimation method, the confidence bounds of these statistics in normal conditions (abnormality-free) can be established as the spatial references. These unique two references will guarantee the reliable spatial localization of abnormality.
Systemic simulations and experiments have been conducted to validate the effectiveness and efficiency of all proposed methods.
| Date of Award | 5 Aug 2020 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Hanxiong LI (Supervisor) |