Identification of a process with control valve stiction using a fuzzy system: A data-driven approach

Many researchers focus on detecting and modelling the valve stiction because it has undesirable effects on the control loop performance, which consequently results in poor product quality and increased energy consumption. It is difficult to model a process with a sticky valve using the mathematical definition because of its nonlinear properties such as stiction, hysteresis, dead band and dead zone. This work aims to develop and determine the appropriate model of a process with stiction, which can be used in controller design to mitigate the undesirable effect of the stiction. To achieve this goal by mapping the process with valve stiction to a fuzzy system, a dynamic fuzzy model of the plant is derived through an iterative well-developed fuzzy clustering algorithm, which generates suitable antecedent parameters from a set of input–output measurements that are obtained from the control output (OP) and the process output (PV). To determine the consequent parameters, the least square (LS) estimation is applied. The results reveal that the obtained data-driven Takagi–Sugeno-type (TS) fuzzy rule-based model can effectively represent an appropriate model of the process with stiction for different amounts of stiction that are obtained from the simulation and different industrial loops.     

An improved valve stiction simulation model based on ISA standard tests

The Choudhury valve model is a widely adopted data-driven model to study the behaviour of valve stiction. A recent study (Garcia, 2008) revealed that valve stiction simulation based on Choudhury’s simulation model (Choudhury, Thornhill, & Shah, 2005) fails to pass eight out of fifteen Industry Standard Architecture (ISA) standard tests (ISA, 2000, ISA, 2006) for real control valves. In this study, the ISA testing results are further elaborated for this model. It is found that three minor deficiencies lead to the discrepancies between the Choudhury Model outputs and the expected ones when (i) the valve input signal changes the direction of travel, (ii) the initial stem position does not stay on the working curves l₁ and l₂, and (iii) the valve input signal changes in a ramp–pause–ramp manner. To address the above deficiencies, an improved version of the Choudhury Model, termed as XCH Model, is proposed. Assessments along with the ISA standards presented by Garcia (2008) demonstrate the proposed XCH Model passes all the ISA standard tests and thus provides a more realistic simulation of a real industrial valve being able to exhibit stiction behaviour.     

Detection and quantification of valve stiction by the method of unknown input estimation

Valve stiction is one of the most common causes of oscillations in industrial process control loops. Such oscillations can degrade the overall performance of the loop and eventually the final product quality. The detection and quantification of valve stiction in industrial process control loops is thus important. From previous studies in the literature, a sticky valve has been shown to have a distinct signature of the stiction phenomena in its valve positioner data. However, the position of the modulating control valves is seldom available. We consider the problem of estimating the valve position as an unknown input estimation problem. In this work, we propose a novel application of the unknown input estimator in order to estimate the valve position given the process model and the data of the process variable and controller output. Using the estimated valve position, we can detect and also quantify the amount of stiction. We demonstrate the efficacy of the method through simulation examples where a sticky valve is deliberately introduced in the closed loop using a two-parameter stiction model available in the literature. Application of the proposed methodology to a laboratory scale flow control loop is presented. An industrial case study is also presented in which the algorithm accurately detects and quantifies stiction in the level control loop of a power plant.     

A simple model-free butterfly shape-based detection (BSD) method integrated with deep learning CNN for valve stiction detection and quantification

Control valve stiction is a long-standing problem within process industries. In most methods for shape-based stiction detection, they rely heavily on the traditional controller output (OP) and process variable (PV) plot (i.e. PV-OP plot) that tends to produce an “elliptical” shape which is the widely acknowledged pattern indication for the presence of stiction. However, many of the methods suffered from unsatisfactory generalization capability when subjected to different loop dynamics. In this paper, a “butterfly” shape derived from the manipulation of the standard PV and OP data, which is more robust towards different loop dynamics, is developed for stiction detection. This simple model-free butterfly shape-based detection (BSD) method uses Stenman's one parameter stiction model, which results in a distinctive ‘butterfly’ pattern in the presence of stiction. The proposed BSD is tested on simulated data, as well as 26 benchmark industrial case studies and has shown a relatively higher generalization capability with relatively higher successful detection rate on stiction loops and on non-stiction loops. A simple quantification algorithm based on BSD-convolutional neural network (BSD-CNN) framework is then developed to quantify the stiction severity. Based on the 15 benchmark industrial loops with stiction, the proposed BSD-CNN quantification algorithm has shown reasonable accuracy when compared to other published quantification methods in literature.     

Estimation of valve stiction in control loops using separable least-squares and global search algorithms

This contribution presents a new procedure for quantifying valve stiction in control loops based on global optimisation. Measurements of the controlled variable (PV) and controller output (OP) are used to estimate the parameters of a Hammerstein system, consisting of a connection of a two-parameter stiction model and a linear low-order process model. As the objective function is non-smooth, gradient-free optimisation algorithms, i.e., pattern search (PS) methods or genetic algorithms (GA), are used for fixing the global minimum of the parameters of the stiction model, subordinated with a least-squares estimator for identifying the linear model parameters. Some approaches for selecting the model structure of the linear model part are discussed. Results show that this novel optimisation-based technique recovers accurate and reliable estimates of the stiction model parameters, dead-band plus stick band (S) and slip jump (J), from normal (closed-loop) operating data for self-regulating and integrating processes. The robustness of the proposed approach was proven considering a range of test conditions including different process types, controller settings and measurement noise. Numerous simulation and industrial case studies are described to demonstrate the applicability of the presented techniques for different loops and for different amounts of stiction.     

An autonomous valve stiction detection system based on data characterization

This paper proposes a valve stiction detection system which selects valve stiction detection algorithms based on characterizations of the data. For this purpose, novel data feature indexes are proposed, which quantify the presence of oscillations, mean-nonstationarity, noise and nonlinearities in a given data sequence. The selection is then performed according to the conditions on the index values in which each method can be applied successfully. Finally, the stiction detection decision is given by combining the detection decisions made by the selected methods. The paper ends demonstrating the effectiveness of the proposed valve stiction detection system with benchmark industrial data.     

控制回路中迟滞建模及在线检测

控制阀门中非线性的存在,比如迟滞、死区等,限制了控制回路的性能;大约30%的控制回路振荡是由于阀门的问题,而迟滞是过程工业中所发现的最普遍的阀门问题;尽管已有很多对迟滞现象的理解和建模的尝试,但是仍然缺少一种简单、直观且能比较精确反映真实阀门特性的模型;着重研究阀门迟滞的机理并结合大量的事例,提出了迟滞的模型以及基于系统辨识的在线检测方法,仿真与实际工业应用表明了该方法的有效性与准确性。     

Automatic detection and quantification of stiction in control valves

Stiction is a common problem in spring-diaphragm type valves, which are widely used in the process industry. Although there have been many attempts to understand and detect stiction in control valves, none of the current methods can simultaneously detect and quantify stiction. Conventional invasive methods such as the valve travel test can easily detect stiction, but are expensive and tedious to apply to hundreds of valves to detect stiction. Thus there is a clear need in the process industry for a non-invasive method that can not only detect but also quantify stiction so that the valves that need repair or maintenance can be identified, isolated and repaired. This work describes a model free method that can detect and quantify stiction that may be present in control valves using routine operating data obtained from the process. No additional excitation or experimentation of the plant is required. Over a dozen industrial case studies have demonstrated the wide applicability and practicality of this method as an useful diagnostic aid in control loop performance monitoring.     

Limitations of nonlinear PCA as performed with generic neuralnetworks

Kramer's (1991) nonlinear principal components analysis (NLPCA) neural networks are feedforward autoassociative networks with five layers. The third layer has fewer nodes than the input or output layers. This paper proposes a geometric interpretation for Kramer's method by showing that NLPCA fits a lower-dimensional curve or surface through the training data. The first three layers project observations onto the curve or surface giving scores. The last three layers define the curve or surface. The first three layers are a continuous function, which we show has several implications: NLPCA "projections" are suboptimal producing larger approximation error, NLPCA is unable to model curves and surfaces that intersect themselves, and NLPCA cannot parameterize curves with parameterizations having discontinuous jumps. We establish results on the identification of score values and discuss their implications on interpreting score values. We discuss the relationship between NLPCA and principal curves and surfaces, another nonlinear feature extraction method.     

Compensation of control valve stiction through controller tuning

Despite numerous studies on modeling and detection of static friction (stiction) in control valves, compensation methods for this problem are limited. In this work, a stiction compensation framework is proposed which is based on the oscillation condition introduced in [17]. This condition was used as a tool to predict occurrence and severity of stiction-induced oscillations in control systems. The aim of this paper is to suggest re-tuning guidelines for controllers with regard to the presence of stiction in the control valve, to eliminate or reduce oscillations. A variety of processes and controllers are studied and recommendations are made in order to eliminate the stiction-induced oscillations. For oscillations that cannot be removed, the proposed method will reduce the frequency and magnitude of oscillations. This compensation framework has also been validated using two different pilot-scale experiments with different types of processes and an industrial control system.     

基于规则的阀门非线性特性离散时间仿真模型

在控制阀非线性特性研究中,Choudhury模型得到了广泛应用。但在输入信号突变较大时,模型输出的阀位与实际阀位之间存在一定的偏差。此外,Choudhury模型只考虑了阀门非线性特性中的黏滞特性,而没有考虑间隙特性。在详细分析产生偏差原因的基础上,通过总结阀门输出特性的规律,增设多个描述阀门状态的变量,提出了一种新的基于规则的阀门非线性特性离散时间仿真模型。该模型对Choudhury模型进行了改进,同时包括了间隙特性。仿真表明在各种输入信号情况下,阀门非线性模型能够详细描述阀门的物理实际。同时,该模型能够用于计算机控制系统中阀门非线性特性的模拟。     

Frequency analysis and compensation of valve stiction in cascade control loops

Valve stiction is often a common problem in control loops and stiction induced oscillation is the main cause of poor performance in control systems. Cascade control is extensively applied in process industry as an effective strategy to restrain disturbances and compensate process nonlinearities. In recent years many studies have been performed on the detection and quantification of valve stiction in single feedback control loops. However, there is a lack in developing a mechanism which can analyze stiction induced oscillation in cascade control loops. This work focuses on the frequency analysis of stiction induced oscillations in cascade control loops and proposes a mechanism of oscillation compensation through outer and inner controller tuning. The effect of oscillation compensation by changing control strategies is also discussed. The theoretical analysis is evaluated through simulation examples and a pilot-scale flow-level cascade control experiment.     

System identification applied to stiction quantification in industrial control loops: A comparative study

A comparative study of different models and identification techniques applied to the quantification of valve stiction in industrial control loops is presented in this paper, with the objective of taking into account for the presence of external disturbances. A Hammerstein system is used to model the controlled process (linear block) and the sticky valve (nonlinear block): five different candidates for the linear block and two different candidates for the nonlinear block are evaluated and compared. Two of the five linear models include a nonstationary disturbance term that is estimated along with the input-to-output model, and these extended models are meant to cope with situations in which significant nonzero mean disturbances affect the collected data. The comparison of the different models and identification methods is carried out thoroughly in three steps: simulation, application to pilot plant data and application to industrial loops. In the first two cases (simulation and pilot plant) the specific source of fault (stiction with/without external disturbances) is known and hence a validation of each candidate can be carried out more easily. Nonetheless, each fault case considered in the previous two steps has been found in the application to a large number of datasets collected from industrial loops, and hence the merits and limitations of each candidate have been confirmed. As a result of this study, extended models are proved to be effective when large, time varying disturbances affect the system, whereas conventional (stationary) noise models are more effective elsewhere.     

Stiction – definition, modelling, detection and quantification

Stiction or high static friction is a common problem in spring-diaphragm type control valves, which are widely used in the process industry. Recently, there have been many attempts to understand, define, model and detect stiction in control valves. There are several methods for detecting stiction, but quantification of the actual amount of stiction still remains a challenge. This paper discusses briefly the definition and modelling of stiction. Then it demonstrates a new method to detect and quantify the actual amount of valve stiction using routine operating data. The proposed method is completely data-driven. No additional excitation or experimentation of the plant is required.     

Frequency analysis and experimental validation for stiction phenomenon in multi-loop processes

The existence of static friction, also known as stiction nonlinearity, in a control valve may lead the entire system to limit cycles. Despite numerous studies on modeling, detection and compensation of this phenomenon, lack of a reliable methodology to predict occurrence and properties of such oscillations, particularly in multi-loop settings, is quite sensible. This work focuses on frequency analysis of multi-loop processes oscillating due to stiction. Before using any of the existing stiction models for the analysis, a comparison between existing stiction models and actual lab data is carried out. Derivation of a mathematical representation of the condition, under which stiction-induced oscillations occur in a multi-loop system, is the main achievement of the proposed analysis. This condition enables users to predict and compare the severity of the oscillations, i.e. values of frequencies and magnitudes, in different situations. Results of the theoretical discussion are validated by both simulation and experimental studies.     

A receding–horizon approach to the nonlinear H∞ control problem

The receding–horizon (RH) methodology is extended to the design of a robust controller of H_∞ type for nonlinear systems. Using the nonlinear analogue of the Fake H_∞ algebraic Riccati equation, we derive an inverse optimality result for the RH schemes for which increasing the horizon causes a decrease of the optimal cost function. This inverse optimality result shows that the input–output map of the closed-loop system obtained with the RH control law has a bounded L₂-gain. Robustness properties of the nonlinear H_∞ control law in face of dynamic input uncertainty are considered.     

Friction compensation for a process control valve

A nonlinear locally intelligent actuator design is developed to control a valve independently of the distributed control system. Nonlinear control is implemented through the direct synthesis of a sliding-stem valve model within a nonlinear structure. Input–output linearization with discontinuity smoothing is used to cancel friction nonlinearities as well as to reduce control action chattering. A closed-loop nonlinear Luenberger observer is used to reconstruct an unmeasurable state as well as to provide robust control action in the presence of parametric uncertainty. A model-based fault detector is developed to monitor serious faults such that a warning may be sent to an operator describing the exact nature of the fault. Fault diagnostic approaches are also provided in the form of threshold detection and fault tree analysis. Setpoint tracking results are provided to compare against linear proportional–integral control. The nonlinear controller is shown to outperform linear control set-point tracking measured by integral absolute error (IAE). In conclusion, the advantages of local nonlinear control are discussed.     

基于神经网络的非线性PCA方法

由于普通的主元分析（PCA）方法无法提取数据中的非线性相关特性,本文提出了一种基于神经网络的非线性PCA（NIPCA）方法,不仅提取了高维原始数据的线性信息还能提取非线性信息。在此基础上进一步提出了样本中显著误差及劣点的检测方法,从而支持对其进行合理剔除或是修正,仿真试验表明它能有效地减小误差点对网络训练精度的影响,大大增强了算法的鲁棒性。     

Valve stiction detection through improved pattern recognition using neural networks

A non-invasive method for detecting valves suffering from stiction using multi-layer feed-forward neural networks (NN) is proposed, via a simple class-based diagnosis. The proposed Stiction Detection Network (SDN) uses a transformation of PV (process variable) and OP (controller output) operational data. Verification of the proposed SDN model’s detection accuracy is done through cross-validation with generated samples and benchmarking with various industrial loops. The industrial loop benchmark predictions of the proposed SDN method has a combined accuracy of 78% (75% in predicting stiction, and 81% for non-stiction) in predicting loop condition, matching capabilities of other established methods in accurately predicting realistic industrial loops suffering from stiction, while also being applicable to all types of oscillatory control signals.