Sensor fault diagnosis in a chemical process via RBF neural networks

Radial basis function (RBF) neural networks are investigated here for process fault diagnosis. The use of the output prediction error, between a neural network model and a non-linear dynamic process, as a residual for diagnosing actuator, component and sensor faults is analysed. It is found that this residual for a dependent neural model is less sensitive to sensor faults than actuator or component faults. This is confirmed in experiments for a real, multivariable chemical reactor. A scheme is then developed utilising a semi-independent neural model to generate enhanced residuals for diagnosing the sensor faults in the reactor. A second neural-network classifier is developed to diagnose the sensor faults from the residuals generated, and results are presented to demonstrate the satisfactory detection and isolation of sensor faults achieved using this approach.     

风力发电系统传感器故障诊断

针对非线性风力发电系统,提出了一种基于滑模观测器的传感器故障诊断方法.基于考虑传感器加性故障的非线性动态模型,利用T--S模糊理论建立风力发电系统全局T--S模型,设计模糊T--S系统滑模故障观测器,产生对故障具有敏感性的残差,实现故障检测.通过等价输出控制方法来维持滑模运动,直接获取故障信息,重构传感器故障.最后以三叶片水平轴风力发电系统为例,仿真验证了该方法的有效性与可靠性.     

Sensor selection for fault diagnosis in uncertain systems

ABSTRACT

Finding the cheapest, or smallest, set of sensors such that a specified level of diagnosis performance is maintained is important to decrease cost while controlling performance. Algorithms have been developed to find sets of sensors that make faults detectable and isolable under ideal circumstances. However, due to model uncertainties and measurement noise, different sets of sensors result in different achievable diagnosability performance in practice. In this paper, the sensor selection problem is formulated to ensure that the set of sensors fulfils required performance specifications when model uncertainties and measurement noise are taken into consideration. However, the algorithms for finding the guaranteed global optimal solution are intractable without exhaustive search. To overcome this problem, a greedy stochastic search algorithm is proposed to solve the sensor selection problem. A case study demonstrates the effectiveness of the greedy stochastic search in finding sets close to the global optimum in short computational time.     

基于MSPCA的传感器故障诊断与数据重构

讨论了基于多尺度主元分析的故障传感器数据重构问题。传统的多尺度主元分析方法没有建立故障传感器数据重构模型,在相关传感器信号的所有尺度上建立主元分析模型进行传感器故障诊断的基础上,将主元分析模型的重构结果组合后进行小波逆变换,设计了能够实现故障传感器数据重构的多尺度主元分析模型,从而实现故障传感器的数据重构。最后,利用试车台液氢供应系统的传感器数据仿真了几种典型传感器故障,并对设计模型实现数据重构的实用性和有效性进行了验证。     

Data-driven sensor fault diagnosis systems for linear feedback control loops

This paper develops a sensor fault diagnosis (SFD) scheme for a multi-input and multi-output linear dynamic system under feedback control to identify different types of sensor faults (bias, drift and precision degradation), particularly for the incipient sensor faults. Feedback control, leading to fault propagation and disguised fault rectification, imposes the challenge on the data-driven SFD. With only available output data in closed loop, the proposed scheme comprises two stages of residual generation and residual evaluation. In the residual generation, a data-driven identification of the residual generator for the feedback control system is proposed. One class of parameters in the residual generator are estimated using process delays while another class of parameters describing the output dynamic are derived by the Bayes’ formula. The means and variances control charts of online calculated residuals are made to judge the root cause. Two case studies are performed to illustrate the effectiveness of the proposed method.     

Actuator fault detection and compensation under feedback control

The problem of unknown input estimation and compensation is studied for actuated nonlinear systems with noisy measurements. The proposed solution is based on high-order sliding-mode differentiation and discrete-time optimization technique. Accuracy of the proposed hybrid estimation scheme is evaluated and stability conditions of the compensating mechanism are established. It is shown that the fault detection delay as well as the smallest detectable fault magnitude can be estimated. Efficiency of the proposed approach is demonstrated through oscillatory failure detection and compensation in aircraft surface servo loops.     

基于模式识别的传感器故障诊断

为满足模式识别故障诊断算法的鲁棒性要求,在小波包分解提取特征向量的基础上,提出了有监督模式分类与无监督模式分类相结合的故障诊断方法．利用小波包分解提取各个频带的能量作为特征向量;采用LVQ神经网络作为有监督的模式分类器进行故障诊断;运用无监督的减法聚类方法对新型故障模式进行辨识．最后,通过动力系统管路流量传感器数据对算法进行检验,验证了所提出方法的实用性和有效性．     

Variable-weighted Fisher discriminant analysis for process fault diagnosis

Variable-weighted Fisher discriminant analysis (VW-FDA) is proposed to improve the fault diagnosis performance of the conventional FDA. VW-FDA incorporates the variable weighting into FDA. The variable weighting is used to find out each weight vector for all faults. After all fault data are weighted by the corresponding weight vectors, the summed fault data can be constructed to magnify each fault’s local characteristics. Then, VW-FDA is performed on the summed fault data rather than the original fault data. It is helpful to extract discriminative features from overlapping fault data. Moreover, the partial F-values with the cumulative percent variation are used for exactly variable weighting, which is indispensable to VW-FDA. The proposed approach is applied to Tennessee Eastman process. The results demonstrate that VW-FDA shows better fault diagnosis performance than the conventional FDA.     

Finite-frequency output feedback MPC for Markov jump systems

This paper investigates a general design framework of dynamic output feedback model predictive control (DOFMPC) for Markov jump systems within both time and frequency domain. Such a design with guaranteed H_∞ and quadratic performance is formulated by a standard semi-definite programming (SDP), and it is achieved by employing a special congruence transformation. The SDP condition greatly reduces the computational effort by eliminating bilinear matrix inequalities or equation constraints reported in existing references. Specifically, the H_∞ norm of the transfer function is optimized within three types of frequency ranges on account of generalized Kalman–Yakubovic–Popov (GKYP) lemma. The quadratic index is optimized online via SDP. Finally, we verify the feasibility and effectiveness of the proposed method from both theoretical and practical point of view.     

基于数据和知识的工业过程监视及故障诊断综述

从复杂工业过程所可能具有的过程特性及数据存取过程中引入的数据特性分析出发,综述了具有复杂数据特性的工业过程的多元统计监视方法,并分别讨论了基于数据和基于知识方法进行故障诊断的优势、进展、适用范围及二者结合的可能.最后探讨了这一领域中值得进一步研究的问题和可能的发展方向.     

New nonlinear residual feedback observer for fault diagnosis in nonlinear systems

The increased complexity of plants and the development of sophisticated control systems have necessitated the parallel development of efficient Fault Detection and Isolation (FDI) systems. This paper discusses a model based technique, viz., observers for detecting and isolating parametric and sensor faults. In this paper, a novel diagonal nonlinear residual feedback observer is proposed which is valid for a certain class of nonlinear systems where, subject to other conditions, the state depends nonlinearly on the fault. A number of typical chemical engineering systems can be represented by models of this form. The structure of the observer ensures that the residuals are diagonally affected by the faults. Conditions for exact decoupling of residuals are presented and convergence of the observer in the presence of step faults is proved using Lyapunov like analysis. Multiple observers and a decision logic module are used for FDI when there are un-monitored faults. Results are presented from numerical simulations of an illustrative example and a typical chemical engineering system: a counter-current heat exchanger.     

PCA-SDG based process monitoring and fault diagnosis

Significant research has been done in recent years to use principal component analysis (PCA) for process fault diagnosis. The general approach involves manual interpretation of measured variable contributions to the residual and/or principal components. For a large chemical process, this could be tedious and often impossible. In addition, it hampers the automation of high-level analysis and decision support tasks that require root cause information. In this work, the interpretation of PCA-based contributions is automated using signed digraphs (SDGs). Also, a serious limitation of SDG-based diagnosis – the assumption of a single fault – is overcome by developing a SDG-based multiple fault diagnosis algorithm. The implementation of the PCA-SDG-based fault diagnosis algorithms is done using G2. Its application is illustrated on the Amoco Model IV Fluidized Catalytic Cracking Unit (FCCU).     

Statistical process control based fault detection of CHP units

This paper describes a fault diagnosis method that provides early detection of fouling of the heat recovery system of combined heat and power units. Early detection of fouling build-up is difficult from basic data analysis methods due to limited instrumentation, and a unit can operate for many months with a reduced heat transfer rate before an unplanned shut down. This novel application of statistical process control (SPC) using an estimate of the coolant flow rate, provides advanced warning of fouling build-up and allows significantly increased energy recovery and reduced financial losses from unplanned unit shut down and incorrect fault identification.     

Monitoring nonstationary and dynamic trends for practical process fault diagnosis

This article introduces a revised common trend framework to monitor nonstationary and dynamic trends in industrial processes and shows needs for each improvement on the basis of three application studies. These improvements relate to (i) the extension of the common trend framework to include sets that contain stationary and nonstationary variables, (ii) handling cases where residuals are not drawn from multivariate normal distributions and (iii) the application of the framework to larger variable sets. Existing work does not adequately address these practically important issues. Industrial application studies highlight the needs for (i) the extended framework to model data sets containing stationary and nonstationary variables, (ii) handling statistics that are not based on normally distributed residuals and (iii) the use of Chigira procedure to robustly extract common trends. The extended framework is compared to traditional approaches.     

Robust MPC for systems with output feedback and input saturation

In this work, it is proposed an MPC control algorithm with proved robust stability for systems with model uncertainty and output feedback. It is assumed that the operating strategy is such that system inputs may become saturated at transient or steady state. The developed strategy aims at the case in which the controller performs in the output-tracking scheme following an optimal set point that is provided by an upper optimization layer of the plant control structure. In this case, the optimal operating point usually lies at the boundary of the region where the input is defined. Assuming that the system remains stabilizable in the presence of input saturation, the design of the robust controller is performed off-line and an on-line implementation strategy is proposed. At each sampling step, a sub optimal control law is obtained by combining control configurations that correspond to particular subsets of available manipulated inputs. Stability of the closed-loop system is forced by considering in the off-line step of the controller design, a state contracting restriction for the closed-loop system. To produce an offset free controller and to attend the case of unknown steady state, the method is developed for a state-space model in the incremental form. The method is illustrated with simulation examples extracted from the process industry.     

Sensor deployment for fault diagnosis using a new discrete optimization algorithm

Optimal allocation of the sensor in a wireless sensor network (WSN) is required to have a satisfactory fault diagnosis within the system. In fact, the sensor nodes in the network should be located in an arrangement to maximize the failure diagnosis. In this paper, the sensor deployment optimization to diagnose the distributed failures in a wireless unmanned aerial vehicles (UAVs) network has been studied. In this way, a novel evolutionary optimization algorithm inspired by the gases Brownian and turbulent rotational motion is utilized which is called Discrete Gases Brownian Motion Optimization (DGBMO) algorithm. An integer linear programming (ILP) approach is used to formulate the sensor deployment. Then the sensor deployment optimization is solved by DGBMO as well as generic ILP solvers and Boolean satisfiability-based ILP solvers. The results show that DGBMO is suitable for sensor disposition optimization especially in large-sized UAV networks.     

基于自适应观测器的四旋翼无人飞行器传感器故障诊断方法

针对四旋翼无人飞行器传感器故障诊断问题,提出一种用于四旋翼无人飞行器加速度计和陀螺仪故障同时发生的故障检测与隔离以及故障偏差值估计的非线性诊断方法.首先,在建立飞行器动力学模型和传感器模型的基础上,构建四旋翼无人飞行器传感器故障检测与诊断系统.其次,利用故障观测器完成传感器故障的检测与隔离,基于Laypunov方法设计非线性自适应观测器对未知故障偏差值进行估计.最后,在传感器测量噪声存在的情况下,证明自适应律的稳定性和参数收敛性.实验结果表明,该方法能有效进行传感器的故障检测与隔离,实现对传感器故障偏差的估计与跟踪.     

A hybrid fault diagnosis strategy for chemical process startups

Development of fault detection and diagnosis has been emphasized for industrial processes in order to reduce process downtimes and maintain high quality products with reduced environmental effects. Faults occur more frequently during process startups due to dramatic state variations and tendency of manual operation, and it is therefore vital to diagnose and correct any faults efficiently during process startups. In this paper, a new fault diagnosis method for process startups is developed using on-line dynamic time warping technique in combination with the principal component analysis. SymCure reasoning under the G2 Optegrity is integrated to the strategy so that the method is able to diagnose new faults unknown to historical data. The proposed method was tested on startups of a lab-scale distillation column. Results indicate that it can diagnose both known and unknown faults effectively with improved computational efficiency.     

Internal representation of a process,fault diagnosis and fault correction

This paper reports on an investigation into the relationship between the internal representation of a process on the one hand and on the other, control behaviour when diagnosing and correcting faults. The subjects were 87 process operator trainees, performing certain tasks in a simulated process control situation. Two modes of internal representation are distinguished: a more verbal or abstract mode of the functioning of the process (the mental model) and a more visual or concrete mode of the structure of the process (the mental image). It is concluded that the mental model probably plays an important role in fault correction and in the verification process in diagnosing faults, while the mental image seems to play an important role in the search for information in the process of diagnosis. Some implications for operator training are discussed.