Đani Juričić

A new modified Smith predictor: the concept, design and tuning.

The paper presents a new modified Smith predictor (MSP) for processes with a long time delay. The MSP appears as an extension of the double controller-scheme (DCS) proposed by Tian and Gao. The important feature of the MSP is that the trade-off between disturbance rejection and robustness to variations in process parameters can be adjusted by means of a single free parameter. The main contribution of the paper concerns tuning of the MSP, which relies on a combination of magnitude optimum criterion with process parameterisation based on multiple integrals of the open-loop step response. In a simulation study the performance of the MSP is compared with that of two known controllers for time delay systems, i.e. DCS of Tian and Gao and Hägglunds predictive PI controller. The results show the advantage of the MSP compared to the two other controllers.

Model-based fault detection and isolation for a gas–liquid separation unit☆

Abstract A complete fault detection and isolation system is designed for a gas–liquid separation unit. It involves the determination and identification of grey box models, the design of a model-based residual generator, and finally the evaluation of the residuals via a set of statistical tests. The latter are cumulative sum (CUSUM) tests which are combined in such a way that both fault detection and fault isolation can be achieved. The performance of the resulting diagnosis system, such as missed alarm rate, wrong isolation rate and mean detection delay, are studied via simulations.

Application of Unscented Transformation in Nonlinear System Identification

Abstract This article addresses the problem of system identification of nonlinear dynamical state-space models from input and output data. The problem is tackled by using Expectation Maximization (EM) algorithm for calculating the Maximum Likelihood (ML) estimates of the model parameters. The novelty of the presented algorithm is related to an efficient employment of the Unscented Transformation (UT) in the expectation step, which lowers the number of computations required. This property enables the algorithm to cope with high-dimensional models without a significant increase in computational load. The overall performance of the algorithm is demonstrated using both numerical and real data examples.

Model-Based Diagnosis for a Gas-Liquid Separation Unit

Abstract The design of a diagnostic system based on mathematical models of the plant components is performed for a gas-liquid separation unit. This design includes first modelling of the process by using physical laws governing its behaviour. Hardware redundancy and simple nonlinear observers are used to synthesise a residual generator with 7 outputs. The generalized likelihood ratio (GLR) test allows one to evaluate the residuals. Tuning of the design parameters is discussed, and a reinitialization procedure is proposed in order to distinguish between temporary and permanent changes in the mean of the residuals. Simulations are used to illustrate the choice of the design parameters

Rényi Entropy Based Statistical Complexity Analysis for Gear Fault Prognostics under Variable Load

In this paper we propose an approach for gear fault prognostics in presumably non-stationary and unknown operating conditions. The approach monitors the evolution of the statistical complexity of the generated vibrations envelope vis-a-vis its Renyi entropy. The statistical complexity is obtained through the wavelet coefficients calculated from the generated vibrations. Such an approach allows seamless estimation of the remaining useful life of the monitored drive without any prior information about the operating conditions and no a priori data regarding the physical characteristics of the monitored drive. The effectiveness of the approach was evaluated on experiments monitoring natural gear fault progress under variable load.

A multiple integration tuning method for filtered PID controller

Abstract The magnitude optimum (MO) technique provides non-oscillatory closed-loop response for a large class of process models. However, this technique is based on a transfer function model that requires precise process identification and extensive computations. In the present paper, it is shown that a close relation exists between multiple integrations of the process step response and the MO criterion. Due to this relation, the MO criterion can be more simply achieved by using filtered PID controller.

A Diagnostic System for Vacuum Cleaner Motors

Abstract A prototype of a diagnostic system for quality end-tests of vacuum cleaner motors is presented. The core of the system are four modules for features extraction based on vibration analysis, sound analysis, analysis of commutation and check of parity relations respectively. The symptoms resulting there from are processed by a reasoning module relying on the Transferable Belief Model (TBM). The result of such a comprehensive diagnostic procedure is detection of faulty motors and localisation of tentative fault origins along with accompanying belief measures. Attributes of the system performance and an excerpt from results obtained in an experimental study over a series of motors are provided.

Vibration Based Diagnosis of Distributed Bearing Faults

Bearings are vital components in rotating machinery, being the only one that separate stationary and rotating elements of the machines. Bearings are subjected to degradation in various manners, e.g. scuffing, pitting, brinelling, electro erosion. In the early stage of degradation, faults emerge on the rolling surface locally. These tiny surface defects are hardly visible but do not cause problems in system operation. As time progresses, the process further deteriorates the rolling surface. The damage on the rolling surface widens in size and becomes distributed fault. The aim of this paper is to propose an approach for detection of distributed bearing faults. It is based on assumption that these faults excite particular vibrational pattern that can make them distinguishable from local faults. For this purpose, a lumped parameter model capable of simulating vibrational response of damaged bearing was adopted. The model was extended by incorporating fluctuations in ball diameter. By analyzing envelope spectra of simulated and measured vibrations we show that vibrational patterns produced by localized and distributed faults differ. The main contribution of the paper is a new feature which combines BPFO (ball pass frequency outer) and FTF (fundamental train frequency) component of the envelope spectrum. A simulation study dealing with broad range of damage propagation scenario is performed in order to assess performance of the proposed features. Finally, validation is performed on the set of damaged bearings.

ROBUST DETECTION OF FAULTS IN COMMAND INPUTS

Abstract The problem of robust fault detection in command inputs in the presence of modelling errors in linear time-invariant systems is addressed. The problem leads to a hypothesis test with a test statistic taking form of a weighted quadratic sum of independent normally distributed variables. It is rather difficult to compute the threshold for such a test and the corresponding error probabilities in an exact manner. Instead, the approximative results are derived by assuming assymptotic distribution of the test statistic towards normal distribution. Convergence conditions in case of growing number of summands are presented. Performance of the resulting detector is demonstrated on a simulated DC motor-generator rig.