Pre-Processing Method to Improve Cross-Domain Fault Diagnosis for Bearing

Abstract

Models trained with one system fail to identify other systems accurately because of domain shifts. To perform domain adaptation, numerous studies have been conducted in many fields and have successfully aligned different domains into one domain. The domain shift problem is caused by the difference of distributions between two domains, which is solved by reducing this difference. Source domain data are labeled and used for training the models to extract the features while the target domain data are unlabeled or partially labeled and only used for aligning. Bearings play important roles in rotating machines, so many artificial intelligent models have been developed to diagnose bearings. Bearing diagnosis has also faced a domain shift problem due to various operating conditions such as experimental environment, number of balls, degree of defects, and rotational speed. Cross-domain fault diagnosis has been successfully performed when the systems are the same but operating conditions are different. However, the results are poor when diagnosing different bearing systems because the characteristics of the signals such as specific frequencies depend on the specifications. In this paper, the pre-processing method was used for improving the diagnosis without prior knowledge such as fault frequencies. The signals were first transformed to a common pattern space before entering the models. To develop and to validate the proposed method for different domains, vibration signals measured from two ball-bearing systems (Case Western Reserve University datasets and Paderborn University datasets) were used. One dimensional CNN models were utilized for verification of the proposed method and the results of the models using raw datasets and pre-processed datasets were compared. Even though each of the ball-bearing systems have their own specifications, using the proposed method was very helpful for domain adaptation, and cross-domain fault diagnosis was performed with high accuracy.

Fault Diagnosis of Bearings With the Common-Domain Data

“…Therefore, signal processing is essential to get rid of noise and to transform into the common pattern space to perform cross-domain fault diagnosis and make generalize the AI models. Accordingly, the method developed in [30] was employed, and the results of signal processing to eliminate noise in the raw data and transform the features into a common pattern space are presented. The results of the processed data (CWRU, PU, and Pump) are illustrated in Figure 10, 11, and 12, respectively.…”

“…A study in which cross-domain fault diagnosis was performed after the signal processing of bearing data collected from different domains into a common space has been reported [30]. The CWRU data were moved to the same pattern space using the Paderborn University (PU) data in [30], whereas the PU and Pump data were moved to the same pattern space using the CWRU data in this study. Various filters, such as low-pass and minimum-phase filters were utilized.…”

“…The reduction in noise will help the model find fault characteristics better by making the characteristics of the fault distinct. To address the above challenges, a preprocessing method developed in [30] was applied to the three different systems. Because a model trained with one system cannot extract defect characteristics from other systems accurately, the model should reflect the various characteristics to improve the ability to extract features.…”

“…If the data in various systems are considered as one, the domain of the training data can be expanded, and various characteristics and specifications can be applied to the models without a specific method. Kim et al transformed the signals of different systems into the same pattern space using signal processing, and the models trained with processed data provided excellent diagnostic performance for other systems [30]. Using expertise information, Zheng et al combined signals acquired from different systems and proposed generalized bearing diagnostic models [31].…”

“…(1) insufficient training data with limited types or levels of defects, and (2) fault classification when only normal data from target domain are available. Therefore, a signalprocessing method [30] is applied to transfer the data into a common domain and reduce noise. For the second issue, an artificial intelligence (AI) models are designed to classify bearing defects with only normal data from target domain using common-domain data.…”