Views
7
Downloands
2
Endüstride yaygın olarak kullanılan asenkron motorların tercih edilmesinin nedenleri hesaplı, dayanıklı ve güvenilir olmalarıdır. Asenkron motorların iç bilezik, bilye ve dış bilezik kısımlarımda oluşan rulman hataları en sık karşılaşılan hatalardandır. Bu nedenle, asenkron motorlarının çalışmasının verimini arttırmak için rulman hatalarının erken bir aşamada belirlenmesi oldukça önemlidir. Bu çalışmada, Case Western Reserve University (CWRU) rulman veriseti kullanılarak, asenkron motor rulmanlarının iç bilezik, dış bilezik ve bilye bölgelerinde oluşan hataların titreşim verilerinden yararlanarak otomatik sınıflandırılması için iki yönlü uzun-kısa süreli bellek tipi (IY-UKSB) tipi derin sinir ağları tabanlı bir yöntem önerilmektedir. Çalışmada, normal rulman ve hatalı rulmana ait titreşim verileri 128, 256, 512 ve 1024 gibi farklı boyutlarda pencerelere ayrılarak, anlık ferekans ve sprektral entropi ile özellik çıkarımı sonucunda önerilen IY-UKSB ağının performansı değerlendirilmiştir. Çalışmada normal ve hatalı rulman verilerinden oluşturulan veriseti üzerinde farklı pencere genişliklerinde test kümesi üzerinde IY-UKSB ağının doğruluğunun ortalama %80 civarında kaldığı, buna karşın normal ve hatalı rulman verilerinin sınıflandırılmasında anlık frekans ve spektral entropi ile özellik çıkarımı sonrası IY-UKSB ağının ortalama %99.28 doğruluk, %99.72 duyarlılık ve %97.53 seçicilik skorlarına ulaştığı görülmüştür. Sonuç olarak, önerilen IY-UKSB ağının hatalı ve normal rulman titreşim verilerinin ayrımı için güçlü bir sınıflandırıcı olduğu değerlendirilmiştir.
The reasons for the preference of asynchronous engines that are widely used in the industry are that they are accounted, sustainable and reliable. Asenchronous engines' rolling errors in the internal arms, balls and external arms are one of the most common errors. Therefore, it is very important to identify rolling errors in a early stage to increase the efficiency of the operation of asynchronous engines. In this study, using the rolling data of Case Western Reserve University (CWRU), a method based on deep nerve networks (IY-UKSB) for two-way long-term memory type (IY-UKSB) for the automatic classification, using the vibration data of the errors occurring in the inner arm, external arm and wrist areas of asynchronous motor rollers. In the study, the vibration data of the normal roll and the wrong roll were divided into windows in different sizes, such as 128, 256, 512 and 1024, and the performance of the IY-UKSB network recommended as a result of the feature extraction with instant ferecance and sprektral entropy was assessed. The study found that the IY-UKSB network’s accuracy on the test set on the data created from normal and incorrect rolling data in different window widths was about 80% on average, but the IY-UKSB network’s accuracy on the classification of normal and incorrect rolling data with instant frequency and spectral entropy and character extraction achieved an average of 99.28% accuracy, 99.72% sensitivity and 97.53% selectivity. As a result, the recommended IY-UKSB network has been assessed as a strong classifier for the distinction between the wrong and normal rolling vibration data.
The reasons for the preference of induction motors, which are widely used in the industry, are that they are affordable, durable and reliable. Bearing errors in the inner race, ball and outer race parts of induction motors are the most common errors. Therefore, it is very important to detect bearing faults at an early stage in order to increase the efficiency of operation of induction motors. In this study, using Case Western Reserve University (CWRU) bearing dataset, bi-directional long-short-term memory type (Bi-LSTM) deep neural networks are proposed for automatic classification of faults in the inner race, outer race and ball regions of induction motor bearings on vibration data. In the study, the performance of the proposed Bi-LSTM network is evaluated as a result of feature extraction using instantaneous frequency and spectral entropy, by dividing the vibration data of normal bearing and faulty bearing into windows of different sizes such as 128, 256, 512 and 1024. In the study, the accuracy of the Bi-LSTM network for the test set with different window widths on the dataset created from normal and faulty bearing data is 80% on average, on the other hand, after feature extraction with instantaneous frequency and spectral entropy in the classification of normal and faulty bearing data, the accuracy of Bi-LSTM network is observed 99.28% accuracy, 99.72% sensitivity and 97.53% specifity scores. As a result, the proposed Bi-LSTM network is considered to be a powerful classifier for the separation of faulty and normal bearing vibration data.
Field : Fen Bilimleri ve Matematik; Mühendislik
Journal Type : Uluslararası
Article Detail 
Similar Articles
Download PDF
Journal Information
Listen Paper
Cite
Print This Page
Share
Mendeley
