Routine Operating Data Research Articles

An alternative method for estimating Hurst exponent of control signals based on system dynamics

One of the indices for evaluating the performance of control systems based on the routine operation data is the Hurst index. It is shown that the Hurst index value is very close to that of the minimum variance index which measures the performance of a control loop in terms of minimum variance benchmark. The advantage of Hurst index over minimum variance index is that it does not require the system model and its parameters. The Hurst index is based on the Hurst exponent, which measures the time series correlations. This paper analytically scrutinizes the effect of parameters and dynamics of a stochastic control system on the output Hurst exponent, which helps better comprehend the Hurst index and its relationship with the minimum variance index. For this purpose, the control system output is modeled as an Autoregressive Moving Average (ARMA) process, and the relationship between the parameters of the ARMA process and the output Hurst exponent is examined. Furthermore, the characteristics of the autocorrelation function of a signal under minimum variance conditions are extracted. This paper also proposes a new technique for estimating the Hurst exponent of control signals based on the autocorrelation function. Simulation examples illustrate the efficiency of the proposed approaches.

A one-class support vector machine for detecting valve stiction

In industrial processes, control valve stiction is known to be one of the primary causes for poor control loop performance. Stiction introduces oscillatory behaviour in the process, leading to increased energy consumption, variations in product quality, shortened equipment lifespan and a reduction in overall plant profitability. Several detection algorithms using routine operating data have been developed over the last few decades. However, with the exception of a handful of recent publications, few attempts to apply classical supervised learning techniques have been published thus far. In this work, principal component analysis, linear discriminant analysis and a one-class support vector machine are trained to detect stiction using time series features as input. These features are extracted from the data using the tsfresh package for Python. The training data consists of simulated stiction examples generated using the XCH stiction model as well as other sources of oscillation. The classifier is subsequently benchmarked against closed-loop stiction data collected in an industrial setting, with performance exceeding that of existing methods.

Determining circuit model parameters from operation data for PV system degradation analysis: PVPRO

Physics-based circuit parameters like series and shunt resistance are essential to provide insights into the degradation status of photovoltaic (PV) arrays. However, calculating these parameters typically requires a full current–voltage characteristic (I-V curve), the acquisition of which involves specific measurement devices and costly methods. Thus, I-V curves of the PV system level are often not available. This paper proposes a methodology (PVPRO) to estimate these I-V curve parameters using only operation (string-level DC voltage and current) and weather data (irradiance and temperature). PVPRO first performs multi-stage data pre-processing to remove noisy data. Next, the time-series DC data are used to fit an equivalent circuit single-diode model (SDM) to estimate the circuit parameters by minimizing the differences between the measured and estimated values. In this way, the time evolutions of the SDM parameters are obtained. We evaluate PVPRO on synthetic datasets and find an excellent estimation of both SDM and the key I-V parameters (e.g., open-circuit voltage, short-circuit current, maximum power, etc.) with an average relative error of 0.55%. The performance, especially the extracted degradation rate of parameters, is robust to various measurement noises and the presence of faults. In addition, PVPRO is applied to a 271 kW PV field system. The relative error between the real and estimated operation voltage and current is less than 1%, suggesting that degradation trends are well captured. PVPRO represents a promising open-source tool to extract the time-series degradation trends of key PV parameters from routine operation data.

DBFact: A better approach to calculate the minimum variance control law for nonminimum phase MIMO systems

AbstractThis paper presents the diagonal with Blaschke products factorization (DBFact) approach to factor out multivariable time delay and nonminimum phase zeros from multi‐input multi‐output (MIMO) systems. Based on that, a new output‐order independent minimum variance (MV) control law for MIMO systems is proposed. The DBFact method is a two‐step factorization procedure, relying on the diagonal and Blaschke factorization methods. This method has the advantage of being a direct and non‐iterative procedure. This new factorization approach allows the calculation of an MV control law considering the multivariable time delay as a limiting‐performance factor and the nonminimum phase zeros and their corresponding directions. Based on the proposed MV control law, a performance benchmark is introduced, which can be calculated by the DBFact filters and routine operating data. The DBFact methodology was applied to two control structures of the linear plant model of Linde's heat integrated air separation, in which the MV control law output‐order dependency property and the suitability of the performance benchmark were evaluated. Some results were compared with those obtained by admitting the generalized interactor matrix instead of the DBFact filters. The results show the capability of the DBFact methodology to factor the nonminimum phase terms to provide a reliable MIMO controller performance benchmark and illustrate the importance of considering the nonminimum phase zeros and their actual directionality in the MV control law.

Closed-loop identification of plant and disturbance models based on data-driven generalized minimum variance regulatory control

The paper addresses a closed-loop identification method based on generalized minimum variance (GMV) evaluation. Since the proposed method uses routine operation data, it requires no extra experiment with an external excitation signal. The model parameters of the plant and the disturbance are obtained simultaneously using a single set of input–output data generated by stochastic disturbance. A new variance criterion for closed-loop identification is derived through the conversion of the GMV evaluation function that has originally been developed for data-driven regulatory control. In the conversion, the feedback invariant polynomial, which is estimated from time series analysis of the closed-output signal, plays a key role. The features of the proposed approach lead to bridge closed-loop identification with control performance assessment as well as data-driven controller parameters tuning. The paper proves that the optimization of the proposed criterion results in the unique optimal solution, which corresponds to the true plant and disturbance model parameters. In numerical examples, the proposed method is applied to datasets obtained from a continuous stirred tank reactor (CSTR), which is operated around an unstable steady state. The result illustrates the effectiveness of the proposed closed-loop identification method.

Practical control performance assessment method using Hurst exponents and crossover phenomena

This paper proposes a new Hurst exponent-based method to evaluate the performance of control loops by estimating Minimum Variance Index (MVI). Incorrect estimation of time delays dramatically affects MVI-based performance assessment accuracy. The main advantage of the proposed method over existing methods is that it only requires routine operation data, without any need for prior information of the system model parameters and time delays. Hence, this method is more suitable to be applied on various industrial control processes. Recently, crossover phenomena have been reported in the graphs used to estimate the Hurst exponent of different control signals, which complicate deciding the Hurst exponents values. To address this problem, the crossover phenomenon is investigated for control signals, then the Hurst index is modified based on it. The proposed method is successfully used to assess performance of various types of control loops such as the cascade loop and verified through simulation results.

Data‐driven generalized minimum variance regulatory control using routine operation data

AbstractThis paper provides a new generalized minimum variance (GMV) control using routine operation data. The proposed method achieves disturbance rejection without additional experiments and reference models, which is different from other data‐driven techniques. In this paper, a new data‐driven criterion is proposed for the Box and Jenkins (BJ) model, which is a more general description including the Auto‐Regressive and Moving Average eXogeneous (ARMAX) model. The paper proves that the optimization of the proposed criterion can achieve GMV control. Numerical examples for two different model structures show the validity of the proposed method. In particular, the application to datasets obtained from a continuous stirred tank reactor (CSTR) demonstrates the efficiency of the proposed method.

Comparison of Semirigorous and Empirical Models Derived Using Data Quality Assessment Methods

With the increase in available data and the stricter control requirements for mineral processes, the development of automated methods for data processing and model creation are becoming increasingly important. In this paper, the application of data quality assessment methods for the development of semirigorous and empirical models of a primary milling circuit in a platinum concentrator plant is investigated to determine their validity and how best to handle multivariate input data. The data set used consists of both routine operating data and planned step tests. Applying the data quality assessment method to this data set, it was seen that selecting the appropriate subset of variables for multivariate assessment was difficult. However, it was shown that it was possible to identify regions of sufficient value for modeling. Using the identified data, it was possible to fit empirical linear models and a semirigorous nonlinear model. As expected, models obtained from the routine operating data were, in general, worse than those obtained from the planned step tests. However, using the models obtained from routine operating data as the initial seed models for the automated advanced process control methods would be extremely helpful. Therefore, it can be concluded that the data quality assessment method was able to extract and identify regions sufficient and acceptable for modeling.

Model-plant mismatch detection for cross-directional processes

This paper presents a two-component framework to detect model-plant mismatch (MPM) in cross-directional (CD) processes on paper machines under model-predictive control. First, routine operating data is used for system identification in closed loop; second, a one-class support vector machine (SVM) is trained to predict MPM. The iterative identification method alternates between identifying the finite impulse response coefficients of the spatial and temporal models. It converges, and the parameter estimates are asymptotically consistent. Coefficient estimates drawn from normal operation are used to train a one-class SVM, which then detects model-plant mismatch in subsequent routine operation. This approach applies to routine operating data without requiring external excitations. It can also distinguish mismatches in the process model from changes in the noise model. Examples of CD processes on paper machines are provided to verify the effectiveness of both components.

Data-driven performance assessment of multivariable control loops using a modified Hurst exponent–based index

Control system performance assessment is significant, especially in practical applications. One of the most important indices for the performance assessment of control systems is minimum variance. Calculating the minimum variance index in multivariate systems requires prior knowledge of system parameters and models, and is therefore an obstacle in practical applications. In this article, an index is proposed for the performance assessment of multivariate control loops, evaluating the system performance with the minimum variance criterion and using only the system’s routine operation data. This index can quantify the performance using neither any prior knowledge of system parameters nor the system’s optimal operation data. The proposed index is based on the Hurst exponent, a parameter for measuring correlations in time series data. In this article, detrended fluctuation analysis and rescaled range analysis are used to estimate the Hurst exponents of system outputs. Using a combination of these Hurst exponents, an index is defined for the performance assessment of multivariate systems. The results of simulation examples illustrate that the proposed index can assess the performance efficiently.

Support vector machine approach for model-plant mismatch detection

We develop a model-plant mismatch (MPM) detection strategy based on a novel closed-loop identification approach and one-class support vector machine (SVM) learning technique. With this scheme we can monitor MPM and noise model change separately, thus separating the MPM from noise model changes. Another advantage of this approach is that it is applicable to routine operating data that may lack any external excitation signals. Theoretical analysis of the proposed closed-loop identification is provided in this paper, showing that it can give a consistent parameter estimate for the process model even in the case where a priori knowledge about the true noise model structure is not available. A set of normal operation data with satisfactory performance is collected as the training data. We build SVM models based on process and noise model estimates from training data to predict the occurrence of MPM in the test data. The proposed technique can be applied to both single-input-single-output (SISO) and multi-input-multi-output (MIMO) systems. Two examples from paper machine control are provided to verify the effectiveness of the proposed MPM detection framework.

Data-based approach for feedback–feedforward controller design using closed-loop plant data

Feedforward control of measurable disturbances is a useful complement to feedback control because feedforward control performs control actions before a disturbance response occurs in a process output. In contrast to the typical model-based design approach, a novel data-based method for designing both the feedback and feedforward controllers is presented in this paper. The controller design directly exploits closed-loop plant data and does not require an identification of process and disturbance models. A feedback proportional–integral–derivative controller and feedforward lead-lag compensator are sequentially designed on the basis of the closed-loop response data for a set-point change and for a disturbance input, respectively. Because the controller design process uses plant data integrals, the proposed method is robust against measurement noise. Moreover, the proposed design method can be applied to improve existing underperforming feedback and feedforward controllers using routine closed-loop operating data. Simulation studies demonstrated that the proposed method outperforms existing methods in designing a feedback–feedforward control system.

Detection of model-plant mismatch in closed-loop control system

The performance of model-based control systems depends a lot on the process model quality, hence the process model-plant mismatch is an important factor degrading the control performance. In this paper, a new methodology based on a process model evaluation index is proposed for detecting process model mismatch in closed-loop control systems. The proposed index is the ratio between the variance of the disturbance innovation and that of the model quality variable. The disturbance innovations are estimated from the routine operation data by an orthogonal projection method. The model quality variable can be obtained using the closed-loop data and the disturbance model estimated by adaptive Least absolute shrinkage and selection operator (Lasso) method. When the order of the disturbance model is less than 2 or the process time delay is large enough, no external perturbations are required. Besides, the proposed index is independent of the controller tuning and insensitive to the changes in disturbance model, which indicates that the proposed method can isolate the process model-plant mismatch from other factors affecting the overall control performance. Three systems with proportional integral (PI) controller, linear quadratic (LQ) controller and unconstrained model predictive control (MPC) respectively are presented as examples to verify the effectiveness of the proposed technique. Besides, Tennessee Eastman process shows the proposed method is able to detect process model mismatch of nonlinear systems.

Model-Plant Mismatch Detection with Support Vector Machines

We propose a model-plant mismatch (MPM) detection strategy based on a novel closed-loop identification approach and one-class support vector machine (SVM) learning technique. With this scheme we can monitor MPM and noise model change separately, thus discriminating the MPM from noise model change. Another advantage of this approach is that it is applicable to routine operating data that may lack any external excitations. Theoretical derivations on the closed-loop identification method are provided in this paper, showing that it can furnish a consistent parameter estimate for the process model even in the case where a priori knowledge about the true noise model structure is not available. We build an SVM model based on process and noise model estimates from training data to predict the occurrence of MPM in the testing data. An example on paper machine control is provided to verify the proposed MPM detection framework.

Parameter-based conditions for closed-loop system identifiability of ARX models with routine operating data

Traditionally, closed-loop system identification in the absence of external excitation has focused on determining the identifiability of plant model based on the interplay between the orders of the different polynomials present. However, due to the presence of the controller, it is possible that the system may not be globally identifiable at a given complexity, but may be locally identifiable given certain restrictions or relationships between the individual parameters present in the system. In order to obtain parameter-specific solutions to the problem, many different approaches can be taken. In this paper, the focus will be primarily on an expectation-based analysis of the Fisher information matrix to determine parameter-based constraints on closed-loop identification. Additionally, a method for determining an analytical expression for the expectation operation will be presented. The proposed approach will be illustrated using a first-order autoregressive model with exogenous input controlled by a lead–lag controller. Monte Carlo simulations are used to validate the resulting constraints.

Multivariate Control Loop Performance Assessment With Hurst Exponent and Mahalanobis Distance

A novel data-driven technique for performance assessment of multivariate control loops that takes into account the interactions within the system is proposed. The technique merges the Hurst-exponent-based single-input single-output controller performance index with Mahalanobis distance to devise a multiple-input multiple-output (MIMO) controller performance index. The distinct advantage over the standard minimum variance index and novelty of the proposed approach lies in its ability to quantify the performance of MIMO controller without the knowledge of interactor matrix or system description, which leads to the technique being insensitive to model plant mismatch and easily applicable to nonlinear systems. Only closed-loop routine operating data are required. This new methodology is tested on benchmark systems from the literature and simulation results are presented. Comparison with minimum variance index-based techniques reveals excellent agreement in the trends of both approaches. The results establish the proposed approach as a promising tool for interactor-matrix-independent MIMO control loop performance assessment.

CARMAモデルにおける未知な外乱特性に対するデータ駆動型一般化最小分散制御

This paper proposes a data-driven control parameter tuning method based on generalized minimum variance (GMV) evaluation in regulatory control. The proposed method can perform with no need for plant characteristics nor disturbance ones. For GMV control, this paper introduces a data-driven variance criterion which consists of the input and output data generated by stochastic disturbance. The control parameters are derived by using system parameters as optimization variables. Advantages of the method include applicability to routine operating data, which brings that additive experiments are not required. The efficiency of the method is demonstrated through numerical examples.

Isolation of plant-wide faults using causality detection methods

Isolation of plant-wide faults in large-scale complex systems is particularly challenging. A methodology to detect and isolate faults is proposed, detecting the faulty variables using univariate control charts and the causality information between them to indicate the source. The clustering of faulty variables uses univariate analysis to avoid the smearing effect brought by multivariate analysis. The variable where the fault took place is indicated, handling fault novelties in a very natural manner. The proposed method is discussed and illustrated through its application to the Tennessee Eastman Process and to routine operating data from a thermoelectric power plant.

Detection of Causal Relationships Based on Residual Analysis

The detection of causal interactions between variables from time series data is an important problem in many research areas. Granger causality is a well-known approach that uses prediction error to infer causality. However, the autoregressive models fitted to data usually do not pass model validation tests based on residual analysis, resulting in low causality values that can be inconclusive. The method proposed here fits models for paired combination of all variables and inferences about causality are provided after performing residual analysis. The model order is increased until the autocorrelation test of residual and cross-correlation test of residuals and input provide an answer about causality. The thresholds to decide the existence of causality are provided directly by the data. Higher order multivariate systems are similarly considered and a test to check if causality is direct or indirect is also proposed. The utility of the proposed approach is illustrated by several examples including application on a simulated data set and routine operating data from industry for causality analysis.

Deterministic Performance Assessment and Retuning of Industrial Controllers Based on Routine Operating Data: Applications

Performance assessment and retuning techniques for proportional-integral-derivative (PID) controllers are reviewed in this paper. In particular, we focus on techniques that consider deterministic performance and that use routine operating data (that is, set-point and load disturbance step signals). Simulation and experimental results show that the use of integrals of predefined signals can be effectively employed for the estimation of the process parameters and, therefore, for the comparison of the current controller with a selected benchmark.