Haiyang Hao

Data-driven monitoring for stochastic systems and its application on batch process

Batch processes are characterised by a prescribed processing of raw materials into final products for a finite duration and play an important role in many industrial sectors due to the low-volume and high-value products. Process dynamics and stochastic disturbances are inherent characteristics of batch processes, which cause monitoring of batch processes a challenging problem in practice. To solve this problem, a subspace-aided data-driven approach is presented in this article for batch process monitoring. The advantages of the proposed approach lie in its simple form and its abilities to deal with stochastic disturbances and process dynamics existing in the process. The kernel density estimation, which serves as a non-parametric way of estimating the probability density function, is utilised for threshold calculation. An industrial benchmark of fed-batch penicillin production is finally utilised to verify the effectiveness of the proposed approach.

A Novel Scheme for Key Performance Indicator Prediction and Diagnosis With Application to an Industrial Hot Strip Mill

In this paper, a data-driven scheme of key performance indicator (KPI) prediction and diagnosis is developed for complex industrial processes. For static processes, a KPI prediction and diagnosis approach is proposed in order to improve the prediction performance. In comparison with the standard partial least squares (PLS) method, the alternative approach significantly simplifies the computation procedure. By means of a data-driven realization of the so-called left coprime factorization (LCF) of a process, efficient KPI prediction, and diagnosis algorithms are developed for dynamic processes, respectively, with and without measurable KPIs. The proposed KPI prediction and diagnosis scheme is finally applied to an industrial hot strip mill, and the results demonstrate the effectiveness of the proposed scheme.

A data-driven multiplicative fault diagnosis approach for automation processes

This paper presents a new data-driven method for diagnosing multiplicative key performance degradation in automation processes. Different from the well-established additive fault diagnosis approaches, the proposed method aims at identifying those low-level components which increase the variability of process variables and cause performance degradation. Based on process data, features of multiplicative fault are extracted. To identify the root cause, the impact of fault on each process variable is evaluated in the sense of contribution to performance degradation. Then, a numerical example is used to illustrate the functionalities of the method and Monte-Carlo simulation is performed to demonstrate the effectiveness from the statistical viewpoint. Finally, to show the practical applicability, a case study on the Tennessee Eastman process is presented.

A New Soft-Sensor-Based Process Monitoring Scheme Incorporating Infrequent KPI Measurements

The development of advanced techniques for process monitoring and fault diagnosis using both model-based and data-driven approaches has led to many practical applications. One issue that has not been considered in such applications is the ability to deal with key performance indicators (KPIs) that are only sporadically measured and with significant time delay. Therefore, in this paper, the data-driven design of diagnostic-observer-based process monitoring schemes is extended to include the ability to detect changes given infrequently measured KPIs. The extended diagnostic observer is shown to be stable and hence able to converge to the true value. The proposed method is tested using both Monte Carlo simulations and the Tennessee-Eastman problem. It is shown that although time delay and sampling time increase the detection delay, the overall effect can be mitigated by using a soft sensor. Furthermore, it is shown that the results are not strongly dependent on the sampling time, but do depend on the time delay. Therefore, the proposed soft-sensor-based monitoring scheme can efficiently detect faults even in the absence of direct process information.

A Canonical Variate Analysis Based Process Monitoring Scheme and Benchmark Study

Abstract Principal component analysis (PCA) and Partial least square (PLS) are powerful multivariate statistical tools that have been successfully applied for process monitoring. They are efficient in dimension reduction and are suitable for processing large amount of data. Nevertheless, their application scope is restricted to static processes where the dynamics are ignored. In order to achieve improved monitoring performance for dynamic processes, in this paper, we propose an effective dynamic monitoring scheme based on the canonical variate analysis (CVA) technique. Different from the standard PCA- and PLS-based techniques which rely on mean-extraction for residual generation, the proposed CVA-based scheme takes process dynamics into account as well. The properties of all three methods are then compared in detail and finally, the improvements of the proposed method are demonstrated on the well-accepted Tennessee Eastman benchmark process.

Data-driven quality monitoring and fault detection for multimode nonlinear processes

This paper addresses the problem of quality monitoring and fault detection in nonlinear processes which are working in different operating points. For such processes the statistical model which is obtained from process data is different from one operating point to another, due to nonlinearities and set-point changes. Therefore the classical methods for quality monitoring and fault detection, e.g. partial least squares (PLS), may not be suitable. To this end, a new approach is proposed based on the modeling of nonlinear process as a piecewise linear parameter varying system, considering the behavior of the plant in each operating point as linear time invariant with different parameters in each operating point. The expectation-maximization (EM) algorithm is used to model the process as a finite mixture of Gaussian components and based on the identified model a Bayesian inference strategy is developed to detect the faults which influence the product quality. Finally, the usefulness of the proposed method is demonstrated on a laboratory continuous stirred tank heater (CSTH) setup.

A KPI-related multiplicative fault diagnosis scheme for industrial processes

In this paper, a key performance indicator (KPI) related multiplicative fault diagnosis scheme is proposed for static industrial processes. This scheme is developed for an alternative algorithm to the standard partial least squares (PLS) based process monitoring, where no design parameter like “latent variable number” is involved. Based on both normal and faulty data sets, the multiplicative fault information is firstly estimated. With this knowledge, the most critical low-level control loop/component is further identified. Different from the existing data-driven additive fault diagnosis approaches, this scheme aims to handle the second order statistics, which is of fatal importance for KPI-related fault diagnosis. Finally, an academic example is investigated to illustrate the functionality of this scheme.

An approach for multimode dynamic process monitoring using Bayesian inference

Abstract The problem of data driven design of fault detection system for dynamic processes when the system is subjected to set-point changes as well as model changes due to different operation regimes is studied in this paper. To this end, a new monitoring approach is developed where the finite Gaussian mixture model is firstly identified directly from available historical data. The mixture model contains the mean and covariance information of the data for each component. Then an observer based residual generator is designed by identification of the parity vectors using covariance matrices of the mixture model. A Bayesian approach is further utilized for the purpose of fault detection. The performance and effectiveness of the proposed approach are finally demonstrated on a continuous stirred tank heater (CSTH) benchmark process.

Data-driven design of KPI-related fault-tolerant control system for wind turbines

In this paper, a scheme for an integrated design of fault-tolerant control (FTC) systems for a wind turbine benchmark is proposed, with focus on the overall performance of the system. For that a key performance indicator (KPI) which reflects the economic performance of the system is defined, and the objective of the proposed FTC scheme is to maintain the system KPI in the admissible range in faulty conditions. The basic idea behind this scheme is data-driven design of the proposed fault-tolerant architecture whose core is an observer/residual generator based realization of the Youla parameterization of all stabilizing controllers with an embedded residual generator for fault detection (FD) purpose. The performance and effectiveness of the proposed scheme are demonstrated through the wind turbine benchmark model proposed in [1].

Integration of Residual Evaluation and Threshold Computation into Switched Fault Detection System

Abstract This paper addresses problems related to the designing a fault detection (FD) for switched systems based on fault detection filter (FDF). The main contribution in this paper is to develop a residual evaluation and threshold computation scheme in an integration approach, to achieve an optimal FD for the switched systems. The basic idea behind this scheme is to utilize the available information provided by each local sub-model in the residual generation and evaluation, by constructing weighting factors, that reflect the effect of each local sub-model. In comparison with the standard FD methods, the proposed scheme will lead to a significant improvement in fault detectability for switched systems. The problem is formulated in the well-established LMI optimization techniques with Multiple Laypunov Functions (MLF).