PLS-based EWMA fault detection strategy for process monitoring

Fault detection (FD) and diagnosis in industrial processes is essential to ensure process safety and maintain product quality. Partial least squares (PLS) has been used successfully in process monitoring because it can effectively deal with highly correlated process variables. However, the conventional PLS-based detection metrics, such as the Hotelling's T² and the Q statistics are ill suited to detect small faults because they only use information from the most recent observations. Other univariate statistical monitoring methods, such as the exponentially weighted moving average (EWMA) control scheme, has shown better abilities to detect small faults. However, EWMA can only be used to monitor single variables. Therefore, the main objective of this paper is to combine the advantages of the univariate EWMA and PLS methods to enhance their performances and widen their applicability in practice. The performance of the proposed PLS-based EWMA FD method was compared with that of the conventional PLS FD method through two simulated examples, one using synthetic data and the other using simulated distillation column data. The simulation results clearly show the effectiveness of the proposed method over the conventional PLS, especially in the presence of faults with small magnitudes.     

A sequential probability ratio test (SPRT) to detect changes and process safety monitoring

Detecting anomalies is an important problem that has been widely researched within diverse research areas and application domains. The early detection of faults may help avoid product deterioration, major damage to the machinery itself and damage to human health. This study proposes a robust fault detection method with an Artificial Neural Network-Multi-Layer Perceptron (ANN-MLP) and a statistical module based on Wald's sequential probability ratio test (SPRT). To detect a fault, this method uses the mean and the standard deviation of the residual noise obtained from applying a NARX (Nonlinear Auto-Regressive with eXogenous input) model. To develop the neural network model, the required training and testing data were generated at different operating conditions. To show the effectiveness of the proposed fault detection method, it was tested on a realistic fault of a distillation plant at the laboratory scale.     

A new fault detection method for non-Gaussian process based on robust independent component analysis

Conventional fault detection method based on fast independent component analysis (FastICA) is sensitive to outliers in the modeling data and thus may perform poorly under the adverse effects of outliers. To solve such problem, a new fault detection method for non-Gaussian process based on robust independent component analysis (RobustICA) is proposed in this paper. A RobustICA algorithm which can effectively reduce the effects of outliers is firstly developed to estimate the mixing matrix and extract non-Gaussian feature called independent components (ICs) by robust whitening and robust determination of the maximum non-Gaussian directions. Furthermore, a monitoring statistic for each extracted IC is constructed to detect process faults. Simulations on a simple example of the mixing matrix estimation and a fault detection example in the continuous stirred tank reactor system demonstrate that the RobustICA achieves much higher estimation accuracy for the mixing matrix and the ICs than the commonly used FastICA algorithm, and the RobustICA-based fault detection method outperforms the conventional FastICA-based fault detection method in terms of the fault detection time and fault detection rate.     

Statistical fault detection using PCA-based GLR hypothesis testing

Safe process operation requires effective fault detection (FD) methods that can identify faults in various process parameters. In the absence of a process model, principal component analysis (PCA) has been successfully used as a data-based FD technique for highly correlated process variables. Some of the PCA detection indices include the T² or Q statistics, which have their advantages and disadvantages. When a process model is available, however, the generalized likelihood ratio (GLR) test, which is a statistical hypothesis testing method, has shown good fault detection abilities. In this work, a PCA-based GLR fault detection algorithm is developed to exploit the advantages of the GLR test in the absence of a process model. In fact, PCA is used to provide a modeling framework for the develop fault detection algorithm. The PCA-based GLR fault detection algorithm provides optimal properties by maximizing the detection probability of faults for a given false alarm rate. The performance of the PCA-based GLR fault detection algorithm is illustrated and compared to conventional fault detection methods through two simulated examples, one using synthetic data and the other using simulated continuously stirred tank reactor (CSTR) data. The results of these examples clearly show the effectiveness of the developed algorithm over conventional methods.     

Joint analysis of process and operator performance in chemical process operational safety

Independent studies of case histories by the Health and Safety Commission in the UK and by a Honeywell led industrial consortium world-wide showed that human errors represent the major cause of failure in process plant operation. In contrast to this discovery the majority of previous studies on computer aided systems for fault detection and diagnosis has focused on the process side. This paper presents a methodology, which can involve human factors into the development of systems for automatic identification and diagnosis of abnormal operations and develops methods and techniques that can be used to simultaneously capture, characterise and assess the performance of operators as well as of the process. A joint process–operator simulation platform is developed which is used as a test-bed for carrying out the study. The process part is a simulator, which simulates in high fidelity the dynamic behaviour of the process that is subject to the influence of various disturbances and operators’ interventions. The operator module is developed as a real-time expert system, which emulates operator’s behaviour in interpretation of received signals, and planning and execution of decisions. The interaction between the two modules is managed through an interaction module, which handles the real-time exchange of data using Dynamic Data Exchange. The interaction module also contains the toolkits for analysing the dynamic behaviour of the joint process–operator system. The method and system are illustrated using a simulated case study.     

Fault diagnosis of pipeline and pump unit systems using status coupling analysis

Earlier studies on fault diagnosis of the pipeline and pump unit systems (PPU) relied mainly on independent equipment analyses, which usually lead to false alarms because of the loss of information fusion. The aim of this study is to utilize the status coupling relationship to improve fault detection sensitivity and reduce false alarm rate. A real-time status identification of related equipment step is added between capturing abnormal signals and listing out diagnosis results. For example, when the pipeline pressure fluctuation is found abnormal, a status analysis of pump units is performed immediately, if the pump units are proven to be operational normally, then the pipeline leak alarm is acknowledged valid. The logical reasoning algorithm is used to capture abnormal conditions of pipeline pressures. The pump unit faults are captured by combining information from multiple sources. Field applications show that the proposed method significantly improves the PPU fault detection capability on fault detection sensitive and accuracy.     

A blended hazard identification methodology to support process diagnosis

A novel hazard identification methodology applied to process systems is presented in this paper. This blended hazard identification (BLHAZID) methodology blends two different types of HAZID methods: the function-driven and component-driven approach. The BLHAZID method is based on a conceptual framework called the Functional Systems Framework, which describes structure–function–goal relationships in process systems.The goals of the BLHAZID methodology are to generate outcomes that contain a high coverage of hazards, describe detailed failure causality in process systems and express this knowledge in a structured form for effective reused in subsequent applications, such as fault diagnosis, operator training, design reviews, fault and event tree construction and hazard updates to satisfy major hazard facility requirements.Both the BLHAZID methodology and the Functional Systems Framework were developed with involvement and advice from two major industrial partners. An industrial case study of a benzene saturation unit is presented to illustrate how the BLHAZID methodology operates in practice.     

航空活塞发动机点火强度均衡性测试及防火研究

针对磁电机故障和点火电嘴故障会使发动机失去动力引起飞行事故,高压导线破损放电可能导致动力系统起火并引发飞机火灾等安全隐患,利用紫外探测技术和虚拟仪器技术设计了航空活塞发动机点火强度均衡性测试系统,利用该测试系统完成航空活塞发动机点火系统的安全隐患检测。以四缸活塞发动机正常点火、高压导线破损、点火电嘴积碳等三种点火系统作为测试对象,测试发动机在不同转速下各气缸点火电火花紫外辐射强度,用以表征点火能量强度。实测得到,在不同转速下点火电嘴积碳故障都会导致点火强度降低;高压导线破损漏电会同时导致点火能量损失与点火强度不均衡,并且与转速呈正相关。测试结果可为航空活塞发动机点火系统故障排除及发现可能存在的高压导线破损放电电气火灾隐患点提供技术依据。     

一种新的基于图模型的间歇过程故障监测方法

目前多元统计方法被广泛用于间歇过程故障监测并已经取得了比较好的效果。但是,统计模型的可解释性能比较差,很难直接利用操作人员积累的安全经验。为应对这些不足,提出了一种基于图模型的间歇过程故障监测方法。利用提出的定性建模方法,过程机理及操作人员的安全经验能够方便地表达。利用在线的过程变量数据和正向推理算法推断生产应处于的状态,利用安全知识及时地发现生产异常。当推断的结论与在线测得的结果矛盾或过程变量超过设定的安全限时,给出解释性输出。通过一个间歇反应案例,验证了提出的方法在模型的可解释性和利用安全经验方面的优势。     

基于模糊Petri网的有载分接开关故障诊断方法研究*

为解决分接开关故障诊断以主观经验、缺乏系统化流程以及诊断结果与分接开关实际发生故障存在偏差等问题,依据当前分接开关主要故障分类,提出基于模糊Petri网的有载分接开关故障诊断模型,并结合分接开关典型故障案例,验证模型有效性。研究结果表明:基于模糊Petri网的分接开关故障诊断模型能够有效处理故障概率中不确定性因素,具有容错性好、运行效率高等优势,研究结果可为提高分接开关故障诊断的准确性、保障电力系统安全稳定运行提供参考。     

Optimum parameters for fault detection and diagnosis system of batch reaction using multiple neural networks

Batch process usually differs from the continuous process because of its time-varying variables and the process parameters. An early detection and isolation of faults in the process will help to reduce the process upsets and keep it safe and reliable. This paper discusses on the application of multi-layer perceptron neural network in detecting various faults in batch chemical reactor based on an esterification process that involves the reaction of ethanol and acetic acid catalyzed by sulfuric acid. A multi-layer feed forward neural network with double hidden layers has been used in the neural network architecture. The detection was based on the different patterns generated between normal and faulty conditions. An optimum network configuration was found when the network produced the minimal error with respect to the training, testing and data validation.     

基于免疫机理的氯化工艺事故风险模糊评价

为预防氯化工艺安全生产事故,基于生物免疫机理,先应用仿生学方法分析了氯化安全生产事故预警体系与生物免疫系统的相似性,结合氯化生产工艺的实际情况,构建了基于抗原-抗体模式的安全生产预警体系的层次模型,并确定了各评价指标的权重。结果表明,影响较大的抗原A指标有A12(个体的防护水平)、A24(氯的毒性)、A34(暴露于危险环境);影响较大的抗体B指标有B12(应急机制的完善)、B26(反应物料比例的控制和连锁)、B29(冷却系统中冷却介质的温度)、B30(冷却系统中冷却介质的压力)。基于所构建的氯化工艺安全生产预警指标体系,结合某厂氯苯生产工艺及操作控制指标,进一步运用模糊综合评价方法,对该工艺的事故风险进行了模糊评价。结果表明,以抗原-抗体模式建立的事故风险模糊综合评价方法可操作性强、效果较好,可以提高事故的预防和控制能力,可在工艺过程的安全综合评价中广泛应用。     

基于Petri网的750kV变电站故障恢复

对750 kV变电站可能发生的常见故障进行详细分类,并据此建立故障恢复推理模型,以便发生故障时能及时快速恢复供电。通过分析750 k V变电站拓扑结构,采用故障分类和基于Petri网建模的方法,将所有可能发生的常见故障分为母线、线路及变压器3类。变电站发生故障时,先根据变电站故障诊断结果,再从故障点隔离到恢复非故障失电区域并建立Petri网模型。故障恢复主要对开关进行操作,因此首先建立开关打开和关断的基本Petri网恢复模块,接着从变压器运行方式及裕量情况给出变压器故障恢复的Petri网模型,然后讨论母线和线路故障时不同情况下的故障恢复Petri网模型,最后根据模型推理过程中点火的变迁,将隔离故障点和恢复供电的断路器记录下来,按照Petri网运行的顺序即可给出最终的恢复方案。通过对750 k V变电站各类复杂故障的建模和推理及由此给出的恢复方案,能够及时给出恢复的操作顺序,方便运行人员快速检修。最后通过HPSim软件对代表性故障进行仿真运行,验证该方法正确、可行且及时。     

燃煤电厂袋式除尘专家系统开发研究

开发了一种袋式除尘系统故障专家诊断方法,用于诊断整个袋式除尘系统的故障现象,通过现场数据采集、操作人员与专家系统的人机对话对故障现象进行分析、推理,并做出相应的解决方案指导操作人员排查故障.燃煤电厂袋式除尘专家系统以知识库、推理机为核心,实现整个袋式除尘系统的设备故障诊断功能和维修故障指导功能,辅以解释机构,人机界面来指导用户如何准确无误地操作运行本系统和袋式除尘控制系统.系统被划分为4个模块:专家系统简介模块,在线故障诊断模块,离线故障诊断模块和专家指导模块.     

基于随机森林的阻变型通风网络故障诊断方法*

为了在矿井通风网络发生阻变型故障时,能够快速准确判断出故障位置和故障量,提出1种基于随机森林的通风网络故障位置和故障量诊断方法。利用矿井通风仿真系统IMVS将唐安矿模拟故障生成空间数据集并进行数据预处理,构建基于随机森林的故障诊断模型,并利用该诊断模型对唐安矿矿井通风网络模拟故障位置和故障量进行判断和预测。引用多种方法对模型进行度量,通过唐安矿模拟实验验证基于随机森林的故障诊断模型的有效性。将随机森林和决策树的故障诊断准确率进行对比,研究结果表明:随机森林较决策树故障准确率有进一步的提高,并发现故障地点失误诊断多是相邻巷道,在一定程度上工作人员对故障地点的判断并不受其影响。     

A hybrid approach to faults detection and diagnosis in batch and semi-batch reactors by using EKF and neural network classifier

This work deals with a new hybrid approach for the detection and diagnosis of faults in different parts of fed-batch and batch reactors. In this paper, the fault detection method is based on the using of Extended Kalman Filter (EKF) and statistical test. The EKF is used to estimate on-line in added to the state of reactor the overall heat transfer coefficient (U). The diagnosis method is based on a probabilistic neural network classifier. The Inputs of the probabilistic classifier are the input–output measurements of reactor and the parameter U estimated by EKF, while the outputs of the classifier are fault types in reactor. This new approach is illustrated for simulated as well as experimental data sets using two cases of reactions: the first is the oxidation of sodium thiosulfate by hydrogen peroxide and the second is alkaline hydrolyse of ethyl benzoate in homogeneous hydro-alcoholic. Finally, the combination of the estimated parameter U using EKF and probabilistic neural network classifier provided the best results. These results show the performance of the proposed approach to monitoring the semi-batch and batch reactors.     

A structural decomposition-based diagnosis method for dynamic process systems using HAZID information

A novel diagnosis method is proposed in this paper that uses the results of the blended HAZID analysis extended to the dynamic case of process systems controlled by operational procedures. The algorithm is capable of finding fault root causes in process systems using nominal and observed possible faulty operational procedure execution traces. The algorithm uses the structural decomposition of the process system and its component-level dynamic HAZID (P-HAZID) tables and executes the diagnosis component-wise by first decomposing the observed execution traces, and then assembling the diagnosis results. The exact structure of the algorithm is also discussed, followed by two case studies on which its operation is demonstrated.     

用于生产过程的故障诊断专家系统

讨论了利用人工智能技术来实现生产过程故障检测与诊断的原理。以液位控制系统为例，说明了“基于规则”和“规则架＋ 规则体”两种方法建立专家系统的过程及知识表达的思路     

起重机回转支承故障演化过程探讨

为深入分析起重机回转支承故障形成机理,从运动、受力、材料以及制造工艺流程四个方面分析了起重机回转支承故障的影响因素,按照故障形成原因及故障发生部位对起重机回转支承常见故障进行分类,并指出这些常见故障的早期形式主要表现为微裂纹和轻度磨损两类,在此基础上对连接螺栓、滚道与滚动体、齿圈常见故障的演化发展过程进行了系统的分析和研究.起重机回转支承常见故障的演化发展受到多个因素的影响,演化过程复杂且具有动态变化的特征,对故障演化过程进行深入的分析有助于起重机械事故的预防,具有重要的工程意义.     

Model selection and fault detection approach based on Bayes decision theory: Application to changes detection problem in a distillation column

The fault detection of industrial processes is very important for increasing the safety, reliability and availability of the different components involved in the production scheme. In this paper, a fault detection (FD) method is developed for nonlinear systems. The main contribution consists in the design of this FD scheme through a combination of the Bayes theorem and a neural adaptive black-box identification for such systems. The performance of the proposed fault detection system has been tested on a real plant as a distillation column. The simplicity of the developed neural model of normal condition operation, under all regimes (i.e. steady-state and unsteady state), used in this case is realised by means of a NARX (Nonlinear Auto-Regressive with eXogenous input) model and by an experimental design. To show the effectiveness of proposed fault detection method, it was tested on a realistic fault of a distillation plant of laboratory scale.