Actuator Fault Diagnosis Using Single and Meta-Classification Strategies

. The paper presents the application of various classi(cid:28)cation schemes for actuator fault diagnosis in industrial systems. The main ob-jective of this study is to compare either single or meta-classi(cid:28)cation strategies that can be successfully used as reasoning means in the diagnostic expert system that is realized within the frame of the DIS-ESOR project. The applied research was conducted on the assumption that classic as well as soft computing classi(cid:28)cation methods would be adopted. The comparison study was carried out within the DAMADICS benchmark problem which provides a popular framework for confronting di(cid:27)erent approaches in the development of fault diagnosis systems.


Introduction
The increasing complexity of recent industrial objects makes the issue of fault diagnosis one of the most important directions of research in modern automatic control and robotics [7,24,32]. Technical systems and processes are required to be safely and reliably operated due to the protection of human life and health, the quality of the environment, as well as the economic interests. It is possible to specify numerous areas of interdependence of human and technical means, where safety plays a key role, for instance in aircraft, spaceship, automotive, power or mining industry. The above mentioned factors cause that new developments in control theory such as passive and active fault-tolerant control approaches are more often applied in these areas of the industry [22,17,5]. A special attention is currently paid on the second type of the advanced control methodologies, where fault diagnosis methods hold a critical importance. The present state of the art in the eld of fault diagnosis shows the really need for development of fault diagnosis expert systems. The goal is to elaborate general-purposes systems with multi-domain knowledge representations and multi-inference engines [28,36,9].
Generally, the fault diagnosis can be divided into three steps [18]: fault detection, fault isolation and fault identication. Moreover, each of them can be developed by means of model-free (based on data), model-based and knowledgebased approaches [22]. In this paper the rst approach, where experimental data are exploited was discussed. In this kind of methods data that represents normal and faulty situations can be obtained from historical databases or from simulators as well as laboratory stands. This data is then used to create state classiers and meta-classiers.
The main goal of this paper is to compare dierent classication strategies that can be successfully used as reasoning means in the diagnostic expert system.
The development of the diagnostic expert system shell with multi-domain knowledge representations and multi-inference engines is realized within the frame of the DISESOR project. The DISESOR is an acronym of the decision support system designed for fault diagnosis of machinery and other equipment operating in underground mines as well as for monitoring potential threats that can occur in such kind of industry. The DISESOR system can be used for dierent purposes, e.g. to assess seismic hazard probabilities in the area of the coal mine, to forecast dangerous increase in the methane concentration in the mine shafts, to detect and localize endogenous res, and also to conduct fault diagnostics of machines working in such environment. This study shows the comparison research of the classication schemes for creating fault diagnosis system of the benchmark actuator [2] which was elaborated on the basis of the activity of the DAMADICS

(Development and Application of Methods for Actuator Diagnosis in Industrial
Control Systems) Research Training Network funded by the European Commission. The current paper is a continuation of the research work presented in [20].
The authors taken into account the majority of reviewers' comments and also proposed a new approach for searching proper values of relevant parameters of classiers used to fault diagnosis. The examined methods are planed to be used for designing the engine of the DISESOR system.

Single and meta-classication strategies
There are many types of classiers available in the literature, as well as dierent concepts of using them are introduced [25]. Some examples are methods based on the similarity between objects in the feature space, probabilistic methods or methods which are based on black box models. Generally, the classication problems can be divided into two groups including approaches of supervised and unsupervised machine learning techniques. In the paper, the authors concentrated the attention only on methods belonging to the rst group. Currently, the information fusion and meta-classication problems are recognized as the most important directions of the research in the domain of supervised learning.
The main idea in this approach is the application of simple classiers working together to solve a problem with better results than it can be done by means of single one or more complicated classiers. There are a lot of dierent kinds of information fusion methods, but the most popular are majority voting, weighted voting, boosting, and AdaBoost [25]. On the other hand, meta-classiers are very often used for the same reason that means its eciency is often higher, than the eciency of the best single classier [26].
The current research trends in developing machine learning methods are focused on ideas of improving the general eciency of dierent classication and meta-classication methods. The most important investigations can be found for instance in [30,39,33,38,4,12]. The main directions presented in these studies are concentrated on optimization techniques which are used to tune relevant parameters of the classical methods, e.g. with the use of evolutionary and particle swarm algorithms. A number of results included in the related works show the benets of using these methods. In case of a task of fault detection and isolation the key features of the signals in time or frequency domains are most commonly used. Industrial actuators may be characterized by a very high complexity which aects the large number of measuring signals and their features. Therefore, another approach aimed at improving the eciency of the classier, and often shortening the time of its learning, is to remove irrelevant variables [14]. There are various methods that can be used in this procedure, e.g. forward or backward selection methods, as well as elimination methods based on statistical measures. Another group of methods stands fusion methods such as bagging, boosting, and the development of these concepts that is AdaBoost method [19,40]. These methods are often more eective than simple classiers but also show some drawbacks.
The advanced concepts were developed to take merits and positive aspects of classic methods and to eliminate their limitations [37]. There are also attempts to connect together several dierent methods such as selection of relevant features and usage of boosting into one algorithm [21]. Such approach may lead to the nal result that should be better than the results of the methods applied separately.
3 Model-free fault diagnosis using dierent classication schemes The idea of the well practised model-free fault detection and isolation method is presented in Fig. 1. It can be seen, that faults are detected and distinguished using primary and redundant process variables. In this method two separated classiers must be created. The rst classier uses the subset of process variables (U ∪ Y ) as its input and it is dedicated for generating diagnostic signals (S), whereas the second one has the same set of input variables but its task is to calculate a fault signature (F ). This classier is triggered in case when the diagnostic signal indicates a fault scenario. The proposed method can be viewed as the extension of the most often used model-free fault diagnosis approach, cf.  [22]. The novelty in this study depends on that single and also meta-classiers are automatically tuned in order to obtain the maximum accuracy of fault diagnosis.
Fault detection and isolation algorithms corresponding to the diagram presented in Fig. 1 can be designed using dierent classication methods [18,22,31].
Generally, it is possible to apply so-called classical (e.g. decision trees, k-nearest neighbour, naive Bayes, etc.) or soft computing approaches (e.g. neural networks, bayesian networks, fuzzy systems, neuro-fuzzy systems, etc.). The paper deals with either classic or soft computing methods. In the next part of the article, model-free fault detection and isolation approaches with the use of dierent Fig. 1: A diagram of model-free fault detection and isolation classication schemes are described. As it was mentioned above, these kinds of methods require data (process variables) corresponding to regular (faultless) and faulty states of the system. In this section, dierent variants of three basic concepts with a single classier, meta-classier and a bank of classiers are applied in order to provide the fault detection and isolation system that is directly based on the process variables.

Fault detection schemas
The rst concept of fault detection is presented in Fig. 2 and this is elabo-

Fault isolation schemes
The rst method of fault isolation is comparable to the method that was proposed for the fault detection. It is presented in Fig  The last concept of fault isolation is shown in Fig. 6. The main idea is based on a bank of classiers that are used to calculate degrees of beliefs for specic faults and unknown states of a device. In this case, M single classiers must be created for M faulty states. Each classier is dedicated for one state only (it is used for detection one fault solely). In the next step, all available variables  and meta-classier The engines of fault detection and isolation schemes presented above can be elaborated with the use of well practised classication methods. The classication problem is possible to be solved using many known approaches, however, in this research the following methods are applied: k -nearest neighbour [1], naive Bayes [10], decision tree [27,6], rules induction [11], neural networks [13,15] and support vector machine [16]. Each of these classiers returns a label of a chosen class and the degrees of belief for all predicted classes. The best solution is pointed at the moment when one of the class is characterised by the belief level equal to 1 and the rest of them are equal to 0. It gives us 100% certainty that a new element should be classied as this particular class.

Verication studies
The proposed schemes of fault detection and isolation were implemented using It is an open source software created for solving data mining problems. The verication studies were conducted on data generated using the DAMADICS simulator [3] in order to investigate selected classication schemes. This simulator was elaborated in collaboration of scientists and engineers to simplify the process of evaluating and comparing dierent methods of fault detection and isolation for industrial systems. In the literature there are available several papers where case study results deal with this problem are presented, see e.g. [35,29,23]. The numeric model is used to simulate an electropneumatic valve (Fig. 7) which is a part of the production line in Lublin sugar factory in Poland. The presented model was created and tuned in MATLAB/Simulink R software taking into account the physical phenomena related to the origin of faults in the real actuator system. This simulator was used to generate the following signals of the process variables: CV -process control external signal, P 1 -inlet pressures on valve, P 2 -outlet pressures on valve, X -valve plug displacement, F -main pipeline ow rate, T -liquid temperature, f -fault indicator. All of these signals were normalized to the range between 0 and 1. In rst fault detection scheme the accuracy of most used classiers is high (above 0, 93). The accuracy of neural net and naive bayes based classiers is a little bit lower than other ones but it is still in the acceptable range. The sensitivity of faultless state detection is very close to 1, 00, which means that almost all samples corresponding to faultless state were classied correctly to this class. It is easy to distinguish faults which are low-correlated with faultless state      The second method presented in Fig. 5 uses six classiers as in the previous method but the outputs of these classiers are connected to a meta-classier.  The results obtained for the meta-classier are compared in Tab. 5 and Tab. 6.
The results obtained for the second scheme of fault isolation (Fig. 5), which was based on the meta-classier were more similar to each other than in the rst case (Fig. 4).           In this study, the authors used a confusion matrix in order to evaluate fault diagnosis systems that were created applying dierent classication schemes.
Nevertheless, the accuracy, sensitivity and precision obtained from a confusion matrix can be directly compared with false and true detection/isolation rates proposed by the authors of the DAMADICS simulator [2]. The results of fault detection and isolation using single or meta-classication strategies that were achieved in this study are comparable to even more advanced methods described in the literature [8] [34]. Furthermore, in this study the whole set of potential faults were investigated, whereas in the related papers only selected states were taken into consideration.

Conclusion
In the paper the application of selected classication schemes for fault diagnosis of the actuator systems was presented. The main purpose of the paper was to compare single and meta-classication strategies that could be successfully used as knowledge representation in the diagnostic expert system that is realized within the frame of the DISESOR project. The research was carried out basing on the well-practised hard and soft computing classication methods. The current paper can be viewed as the extension of the research work presented in [20]. In this study, the authors proposed a new approach for searching proper values of relevant parameters of classiers used to fault diagnosis.
The examined methods were tested in the context to be used for designing the engine of the DISESOR system. The comparison study was carried out within the DAMADICS benchmark problem. The classication schemes were implemented in RapidMiner software which is a well-known open source system for data mining and knowledge discovery. The particular results of the fault detection procedures showed that for simple industrial actuators it is possible to apply simple classication schemes without the necessity of using more advanced methods which are based on meta-classiers. The nal results reached in this paper are much better than results showed in [20]. The features and parameters of classiers can be automatically tuned to increase their accuracy and sensitivity.
Overall, the application of single or meta-classication strategies with optimizing of relevant parameters allows to create eective as well as relatively less-complicated computational fault detection and isolation systems that can be successfully employed for on-line and o-line fault diagnosis of industrial actuators.