Ricardo Dunia

Joint diagnosis of process and sensor faults using principal component analysis

Abstract This paper presents a unified approach to process and sensor fault detection, identification, and reconstruction via principal component analysis. The principal component analysis model partitions the measurement space into a principal component subspace where normal variation occurs, and a residual subspace that faults may occupy. Both process faults and sensor faults are characterized by a direction vector, which describes the behavior of the fault. Fault reconstruction is accomplished by sliding the sample vector as close as possible to the principal component subspace. When the actual fault is assumed, the maximum reduction in the squared prediction error is achieved. A fault-identification index is defined in terms of the reconstructed squared prediction error. Fault detectability, reconstructability, and identifiability conditions are derived and demonstrated with a geometric interpretation. Numerous examples are provided to verify the method and conditions derived in the paper. An unreconstructed variance is defined and used to determine the number of principal components for best reconstruction. The proposed approach is applied to a data set from an industrial boiler process.

Determining the number of principal components for best reconstruction

Abstract A well-defined variance of reconstruction error (VRE) is proposed to determine the number of principal components in a PCA model for best reconstruction. Unlike most other methods in the literature, this proposed VRE method has a guaranteed minimum over the number of PCs corresponding to the best reconstruction. Therefore, it avoids the arbitrariness of other methods with monotonic indices. The VRE can also be used to remove variables that are little correlated with others and cannot be reliably reconstructed from the correlation-based PCA model. The effectiveness of this method is demonstrated with a simulated process.

A unified geometric approach to process and sensor fault identification and reconstruction: the unidimensional fault case

Fault detection and process monitoring using principal component analysis (PCA) have been studied intensively and applied to industrial processes. This paper addresses some fundamental issues in detecting and identifying faults. We give conditions for detectability, reconstructability, and identifiability of faults described by fault direction vectors. Such vectors can represent process as well as sensor faults using a unified geometric approach. Measurement reconstruction is used for fault identification, and consists of sliding the sample vector towards the PCA model along the fault direction. An unreconstructed variance is defined and used to determine the number of principal components for best fault identification and reconstruction. The proposed approach is demonstrated with data from a simulated process plant. Future directions on how to incorporate dynamics and multidimensional faults are discussed.

Use of principal component analysis for sensor fault identification

This paper make use of PCA for sensor fault identification via reconstruction. The principal component model captures the measurement correlations and reconstructs each variable to define associated residuals and a Sensor Validity Index (SVI). The filter applied to the SVI adds an important feature for sensor fault isolation because reduces the effect of false alarms and allows the identification of different types of sensor faults.

Determining the Number of Principal Components for Best Reconstruction

Abstract A well-defined unreconstructed variance (URV) is proposed to determine the number of principal components in a PCA model for best reconstruction. Unlike most other methods in the literature, this proposed URV method has a guaranteed minimum over the number of PC’s corresponding to the best reconstruction. Therefore, it avoids the arbitrariness of other methods with myotonie indices. The URV can also be used to remove variables that are little correlated with others and cannot be reliably reconstructed from the correlation-based PCA model. The effectiveness of this method is demonstrated with a simulated process.

Sensor Fault Identification and Reconstruction Using Principal Component Analysis

Abstract This paper presents the use of Principal Component Analysis (PCA) for sensor fault identification via reconstruction. The principal component model captures the measurement correlations and reconstructs each variable by using iterative substitution and optimization. The effect of different sensor faults on model based residuals is analyzed and a new indicator called the Sensor Validity Index (SVI) is defined to determine the status of each sensor. An example using boiler process data demonstrates the attractive features of the SVI.

Model predictive control of remotely operated underwater vehicles

This paper describes the implementation of a model predictive controller novel in an underwater robot vehicle. This work also shows the development of an underwater vehicle model that accounts for physical, hydrodynamic and restorative effects, while the damping coefficients are neglected in the prediction of the vehicle position and orientation. The vehicle kinematic and dynamic models are linearized and arranged into the state space form inside the predictive controller. The model helps to determine the future position and orientation of the vehicle to track a predefined underwater trajectory in an optimal way. The results show that the predictive controller offered significant benefits compared to PID controllers by reducing the MSE and RMS by 40% and 76% respectively.

Effect of process uncertainties on generic model control: a geometric approach

Abstract Techniques available in the literature to enhance the robustness of generic model control (GMC) are reviewed. We show that none of these techniques utilize a rigorous analysis of the process-model mismatch. This paper presents GMC from the geometric perspective in order to analyze the effect of process uncertainties in the closed-loop response. The geometric approach defines a performance surface for the closed-loop response with no model error. When the model is not perfect, the system leaves the surface, making the response diffrent from the reference. A technique based on sliding mode control, called sliding generic model control, is developed here to take into account the process uncertainties and to keep the system close to the GMC performance surface.

Nonlinear model-based fault detection with fuzzy set fault isolation

This paper presents a nonlinear fault detection and isolation system that is able to distinguish single faults that have the same fault signatures. The detection mechanism is based on nonlinear state estimation. Fuzzy set theory followed by parameter estimation of certain parameters of the fault-free model are applied for fault isolation. This parameter estimation step is used to differentiate between a variety of faults, including those with similar signatures. The proposed fault detection and isolation (FDI) method is validated using an air heater lab experiment. Actuator and sensor faults are considered and comparisons with other methods are presented and analyzed under different fault scenarios. The proposed FDI method shows significant advantages when it is applied to nonlinear model systems with fault-free models available.

Graphical based predictive control design

Graphical programming provides a suitable way to design and configure advanced control applications. This paper illustrates the configuration of graphical model predictive control applications. The system in study is controlled with linear as well as nonlinear predictive control methods, and it demonstrates the convenience of a graphical interface for the development of these popular control design techniques. The system built uses the LabVIEW identification, control design optimisation and simulation toolkits for design verification and deployment. The control solutions can be easily imported to a real time platform for industrial applications. The predictive control with constrains is presented in details in this paper and its performance is illustrated in a case study. The example shows how graphical interface control design increases the reliability and robustness of the controller implementation