This paper deals with the early detection of fault conditions in induction motors using a combined model- and machine-learning-based approach with flexible adaptation to individual motors. The method is based on analytical modeling in the form of a multiple coupled circuit model and a feedforward neural network. In addition, the differential evolution algorithm independently identifies the parameters of the motor for the multiple coupled circuit model based on easily obtained measurement data from a healthy state. With the identified parameters, the multiple coupled circuit model is used to perform dynamic simulations of the various fault cases of the specific induction motor. The simulation data set of the stator currents is used to train the neural network for classification of different stator, rotor, mechanical, and voltage supply faults. Finally, the combined method is successfully validated with measured data of faults in an induction motor, proving the transferability of the simulation-trained neural network to a real environment. Neglecting bearing faults, the fault cases from the validation data are classified with an accuracy of 94.81%.

Condition monitoring of machines and powertrain components is an essential part of ensuring reliability and product safety in many industries. The monitored machines and components are often divided into different condition classes as well as classified using machine learning methods. In order to enable classification with machine learning algorithms, the acquisition of a sufficient amount of data from each condition class is essential. In reality, the collection of data for faulty system states turns out to be much more difficult, therefore in many use cases balanced data sets are not available. However, when classifying faulty states, an identical number of data per class is of great importance. This problem can be counteracted with synthetic data generation. Generative Adversarial Networks (GAN) are a suitable approach to generate synthetic data based on real measured data. In most cases of synthetic data generation, different damage cases, e.g. from a transmission, are simulated, but a generation of synthetic data is not performed at different operating conditions. However, different speeds and torques are a reality when monitoring, as the drive systems operate under changing operating conditions. Therefore, in the context of this paper, synthetic data generation at different operating states is investigated in order to implement a condition monitoring system for good and bad system conditions which includes different operating states. So, vibration data is acquired at different operating conditions of a transmission on a drive test rig and relevant features are highlighted using a suitable signal pre-processing method. The features, caused by different operating conditions, can also be generated synthetically by GAN. Therefore, it is possible to achieve a similar classification accuracy by integrating synthetically generated data as with real data, which makes the synthetic data generation a viable solution for extending existing data sets.

In many areas of drive technology, condition monitoring of transmissions and drive systems is becoming an increasingly important discipline. Condition monitoring systems are used in many cases in combination with machine learning algorithms. The generation of a sufficient amount of data per condition class is relevant to ensure training stability and accuracy of the applied algorithms. Especially in early development phases a sufficient data generation is not often given. In the scope of this paper, a Generative Adversarial Network is applied to generate synthetic data and therefore extend existing measurement data sets. Acceleration data in three different condition classes is used, that has been collected on a gearbox as part of the PHM Data Challenge 2009. In order to highlight relevant features and reduce the number of data points, data is pre-processed via appropriate signal analysis techniques, in this case with the spectral kurtosis. It is shown, that in this use case the synthetically generated data via a Generative Adversarial Network has the same feature characteristics as the real measured data sets. The augmentation of the existing data set also improves the detection accuracy with artificial neural networks for the classification of different system states.

No abstract

Purpose Monitoring and diagnosis of fault cases for squirrel cage induction motors can be implemented using the multiple coupled circuit model. However, the identification of the associated model parameters for a specific machine is problematic. Up to now, the main options are measurement and test procedures or the use of finite element method analyses. However, these approaches are very costly and not suitable for use in an industrial application. The purpose of this paper is a practical parameter identification based on optimization methods and a comparison of different algorithms for this task. Design/methodology/approach Population-based metaheuristics are used to determine the parameters for the multiple coupled circuit model. For this purpose, a search space for the required parameters is defined without an elaborate analytical approach. Subsequently, a genetic algorithm, the differential evolution algorithm and particle swarm optimization are tested and compared. The algorithms use the weighted mean squared error (MSE) between the real measured data of stator currents as well as speed and the simulation results of the model as a fitness function. Findings The results of the parameter identification show that the applied methodology generally works and all three optimization algorithms fulfill the task. The differential evolution algorithm performs best, with a weighted MSE of 2.62, the lowest error after 1,000 simulations. In addition, this algorithm achieves the lowest overall error of all algorithms after only 740 simulations. The determined parameters do not completely match the parameters of the real machine, but still result in a very good reproduction of the dynamic behavior of the induction motor with squirrel cage. Originality/value The value of the presented method lies in the application of condition-based maintenance of electric drives in the industry, which is performed based on the multiple coupled circuit model. With a parameterized model, various healthy as well as faulty states can be calculated and thus, in the future, monitoring and diagnosis of faults of the respective motor can be performed. Essential for this, however, are the parameters adapted to the respective machine. With the described method, an automated parameter identification can be realized without great effort as a basis for an intelligent and condition-oriented maintenance.