Giulio Gola

Feature-based classifier ensembles for diagnosing multiple faults in rotating machinery

Recent researches in fault classification have shown the importance of accurately selecting the features that have to be used as inputs to the diagnostic model. In this work, a multi-objective genetic algorithm (MOGA) is considered for the feature selection phase. Then, two different techniques for using the selected features to develop the fault classification model are compared: a single classifier based on the feature subset with the best classification performance and an ensemble of classifiers working on different feature subsets. The motivation for developing ensembles of classifiers is that they can achieve higher accuracies than single classifiers. An important issue for an ensemble to be effective is the diversity in the predictions of the base classifiers which constitute it, i.e. their capability of erring on different sub-regions of the pattern space. In order to show the benefits of having diverse base classifiers in the ensemble, two different ensembles have been developed: in the first, the base classifiers are constructed on feature subsets found by MOGAs aimed at maximizing the fault classification performance and at minimizing the number of features of the subsets; in the second, diversity among classifiers is added to the MOGA search as the third objective function to maximize. In both cases, a voting technique is used to effectively combine the predictions of the base classifiers to construct the ensemble output. For verification, some numerical experiments are conducted on a case of multiple-fault classification in rotating machinery and the results achieved by the two ensembles are compared with those obtained by a single optimal classifier.

A neuro-fuzzy technique for fault diagnosis and its application to rotating machinery

Malfunctions in machinery are often sources of reduced productivity and increased maintenance costs in various industrial applications. For this reason, machine condition monitoring is being pursued to recognise incipient faults. In this paper, the fault diagnostic problem is tackled within a neuro-fuzzy approach to pattern classification. Besides the primary purpose of a high rate of correct classification, the proposed neuro-fuzzy approach also aims at obtaining an easily interpretable classification model. The efficiency of the approach is verified with respect to a literature problem and then applied to a case of motor bearing fault classification.

Genetic algorithms for signal grouping in sensor validation: A comparison of the filter and wrapper approaches

Sensor validation is aimed at detecting anomalies in sensor operation and reconstructing the correct signals of failed sensors, e.g. by exploiting the information coming from other measured signals. In field applications, the number of signals to be monitored can often become too large to be handled by a single validation and reconstruction model. To overcome this problem, the signals can be subdivided into groups according to specific requirements and a number of validation and reconstruction models can be developed to handle the individual groups. In this paper, multi-objective genetic algorithms (MOGAs) are devised for finding groups of signals bearing the required characteristics for constructing signal validation and reconstruction models based on principal component analysis (PCA). Two approaches are considered for the MOGA search of the signal groups: the filter and wrapper approaches. The former assesses the merits of the groups only from the characteristics of their signals, whereas the latter looks for those groups optimal for building the models actually used to validate and reconstruct the signals. The two approaches are compared with respect to a real case study concerning the validation of 84 signals collected from a Swedish boiling water nuclear power plant.

A randomized model ensemble approach for reconstructing signals from faulty sensors

On-line sensor monitoring aims at detecting anomalies in sensors and reconstructing their correct signals during operation. The techniques used for signal reconstruction are commonly based on auto-associative regression models. In full scale implementations however, the number of sensors to be monitored is often too large to be handled effectively by a single reconstruction model. In this paper we propose to tackle the problem by resorting to a pool (ensemble) of reconstruction models, each one handling an individual group of signals. This approach involves two main technical steps: firstly, a procedure for constructing signal groups, and secondly a procedure for combining the outputs of the reconstruction models associated to the groups. For the signal grouping step, a wrapper optimization search is proposed to identify the optimal number of groups in the ensemble and the size of the groups. For the model output aggregation step, a simple arithmetic average is adopted. Ensemble accuracy and robustness is achieved by promoting diversity between the signal groups through the use of the Random Feature Selection Ensemble (RFSE) technique in combination with the Bootstrapping AGGregatING (BAGGING) technique for training data selection. The individual reconstruction models are based on Principal Components Analysis (PCA). The proposed approach has been applied to a real case study concerning 215 signals monitored at a Finnish nuclear pressurized water reactor. The results obtained have been compared with those achieved by an equivalent ensemble of models based on a grouping directly optimized by a Multi-Objective Genetic Algorithm (MOGA).

Robust nuclear signal reconstruction by a novel ensemble model aggregation procedure

Monitoring of sensor operation is important for detecting anomalies and reconstructing the correct values of the signals measured. This can be done, for example, with the aid of auto-associative regression models. However, in practical applications, difficulties arise because of the need for handling large numbers of signals. To overcome these difficulties, ensembles of reconstruction models can be used. Each model in the ensemble handles a small group of signals and the outcomes of all models are eventually combined to provide the final outcome. In this work, three different methods for aggregating the model outcomes are investigated and a novel procedure is proposed for obtaining robust ensemble-aggregated outputs. Two applications are considered concerning the reconstruction of 920 simulated signals of the Swedish Forsmark-3 Boiling Water Reactor (BWR) and 215 signals measured at the Finnish Pressurised Water Reactor (PWR) situated in Loviisa.

Diagnosing faults in nuclear components by an ensemble of feature-diverse fuzzy classifiers

Ensembles of classifiers offer higher classification accuracy than single classifiers. One method for constructing an ensemble is to have the base classifiers work on different feature sets. In this paper, we present a method for selecting the feature sets of the base classifiers by means of a multi-objective genetic algorithm, aimed at maximising the classification performance and the diversity among the classifiers and at minimising the number of features in the subsets. A static voting technique is used to effectively combine the outputs of the base classifiers to construct the ensemble output. The proposed approach is applied to the classification of (simulated) transients in the feedwater system of a boiling water reactor, and the results are compared with those obtained using an optimal single classifier.

Ensemble of Kernel Regression Models for Assessing the Health State of Choke Valves in Offshore Oil Platforms

AbstractThis paper considers the problem of erosion in choke valves used on offshore oil platforms. A parameter commonly used to assess the valve erosion state is the flow coefficient, which can be analytically calculated as a function of both measured and allocated parameters. Since the allocated parameter estimation is unreliable, the obtained evaluation of the valve erosion level becomes inaccurate and undermines the possibility of achieving good prognostic results. In this work, cluster analysis is used to verify the allocated parameter values and an ensemble of Kernel Regression models is used to correct the valve flow coefficient estimates.

On-Line Fault Recognition System for the Analogic Channels of VVER 1000/400 Nuclear Reactors

In this paper, a method for on-line fault diagnosis of analogic channels of the VVER1000/440 monitoring systems is proposed. The method is based on the analysis of the amplitude fluctuations of electrical signals at the output of analogic channels. The advantage of this method is the possibility to perform on-line fault diagnosis of the monitoring system during the normal operation of the nuclear power plant. The method is also considered to be simple enough for practical implementation and use. The paper presents the practical results of the tests carried out to verify the method with the aim of demonstrating that the shape of the histograms of the amplitude fluctuations characterizes different types of sensor faults and component faults in the analogic channels, such as drifts, frozen or noisy signals. The parameters of the histograms of the amplitude fluctuations are used to construct a fault recognition system which is based on the Pearsons chi-square criterion for verifying the probability hypothesis of the discrete random variable.

A HYBRID ENSEMBLE APPROACH FOR PROCESS PARAMETER ESTIMATION IN OFFSHORE OIL PLATFORMS

The valve flow coefficient is commonly used as a parameter to assess the erosion state of choke valves in offshore oil platforms. In particular, the difference between the theoretical value of the valve flow coefficient and its actual value calculated during operation is retained as the valve health indicator. The actual valve flow coefficient is analytically calculated from the oil, water and gas mass flow rates. These quantities, which are allocated on a daily basis based on the measured total production from a number of wells, on physical parameters (pressures and temperatures) related to the specific well, and on a physical model of the process, can be affected by large uncertainties. Based on such values, the evaluation of the health indicator becomes unreliable and undermines the possibility of using it for prognostic purposes. Similar situations arise every time health monitoring rely on unreliable measurement taken by sensors subject to hard working condition, as often happen in the nuclear industry. This paper proposes a method to obtain more accurate daily estimates of the actual values of the oil, water and gas flow rates, from which improved estimates of the flow coefficient will follow. In this respect, an hybrid ensemble aggregating the physical model with data-driven models built using the Kernel Regression (KR) method has been used. Ensemble diversity is ensured by using different training sets;a local procedure based on the historical performance of the models is adopted to aggregate their predictions. The method is verified on real measurements performed on a number of similar offshore choke valves.

Improving Management and Control of Drilling Operations with Artificial Intelligence

In oil and gas industries, drilling is a complex and critical operation which require constant and accurate real-time monitoring. To this aim, real-time models are required to provide an overview of the drilling operations when direct and reliable measurements are not available. Given the harsh operating environment, sensor reliability and calibration are critical issues and bad data quality is a typical problem which affects the accuracy of the model. As a result, the driller may be misled about the down-hole situation or receive conflicting claims about operating conditions. This paper presents two approaches based on the use of artificial intelligence to improve monitoring of drilling processes in terms of reduced uncertainty and increased confidence. The first exploits the aggregation of the opinion of different experts within a so-called ensemble approach; the second is based on a so-called grey-box approach which combines a physical model and artificial intelligence. The two approaches are applied to the problem of predicting the bottom-hole pressure during a managed pressure drilling operation to demonstrate the improved accuracy and robustness.