Bahman Ahmadi

Condition monitoring plays a vital role in ensuring the safety and reliability of rotating machinery and prevents sudden shutdowns by identifying potential faults before they occur. This process involves collecting, analyzing, and interpreting operational data from machinery to identify abnormal changes in performance that may indicate a fault or impending failure. Among the various condition monitoring methods, vibration signal analysis is considered one of the most effective due to its high sensitivity to mechanical faults and ability to provide real-time insights into machine health. Bearings are one of the most important components of rotating machines, which are also common sources of machine failures and often show signs of failure before other components. Faulty bearings usually produce specific vibration patterns, which makes vibration signal processing a vital tool for timely fault detection and proper machine maintenance. Vibration signals from faulty bearings contain valuable information about the internal health of the machine. Due to the communication and contact of faulty components with other components, these signals produce different vibration patterns compared to the healthy state, which leads to early detection of faults before the machine fails. However, vibration signals in the real world are often contaminated by various noises caused by operating conditions, such as environmental noise or interference noise. These noises can hide fault-related information and lead to incorrect fault detection. Therefore, the use of effective noise reduction techniques is essential to increase the clarity and reliability of vibration signal analysis. In this context, wavelet transform has been proposed as a powerful tool for noise removal. This method is able to decompose the signal into different frequency bands and simultaneously preserve time and frequency information. Unlike conventional methods such as Fourier transform that only provide frequency information, wavelet transform allows the analysis of non-stationary and transient signals. This feature makes wavelet transform a very suitable tool for analyzing signals generated by faulty bearings. Various thresholding methods are used in the noise removal process using wavelet transform to identify and remove the noisy components of the signal. In this study, wavelet transform is used to denoise the vibration signals of faulty bearings. With the aim of improving the accuracy of fault detection and ensuring more accurate condition monitoring, various wavelet functions and thresholding methods are used to evaluate their effectiveness in noise reduction. The performance of denoising techniques is measured using evaluation criteria that do not require a clean signal of the raw signal without noise, such as skewness, cross-correlation coefficient, and signal-to-noise ratio (SNR). In addition, a Shannon entropy-based weighting method is used to determine the effectiveness of the evaluation criteria, which allows for a comprehensive and objective evaluation of the performance of denoising methods in different configurations. This weighting approach ensures that the most valid signal quality indicators receive appropriate emphasis in the analysis process. The results of this study provide valuable insights into the best approaches for wavelet transform-based denoising for vibration signal analysis in condition monitoring. By increasing the clarity of signals related to faulty bearings, these techniques enable more accurate fault detection, minimize the risk of sudden machine failures, and help develop reliable and efficient predictive maintenance strategies for rotating machinery. The findings of this study also demonstrate the importance of appropriate selection of wavelet functions and thresholding methods, and highlight the role of entropy-based weighting in improving the evaluation process. Finally, this study emphasizes the potential of advanced signal processing techniques in transforming condition monitoring methods and ensuring the long-term health and performance of industrial systems.

Rotary machines represent a versatile class of mechanical equipment, used in a wide range of industrial applications, made vital by their high strength and low cost. Bearings are essential components of rotating machines, which have important functions in various applications, since these machines are used for daily operations. The harsh working environment and long-term operation makes them vulnerable to damage, and due to the importance of bearings, it is very important to identify and fix their faults. So far, various methods have been used to detect defects. whose core is the vibration analysis method, the use of a powerful method can affect the quality of the extracted information. Accurately extracting information from small defects is the key to identifying and diagnosing bearing defects. In this thesis, the "experimental wavelet transform" technique, which is a new method for detecting bearing defects at different rotational speeds, is used, and the signal is decomposed into several combinations and experimental modes are obtained using MATLAB software. Also, for the most suitable component, we apply the elongation parameter on the modes. In order to detect the fault, by applying the Hilbert transform on the determined mode, the Hilbert spectrum is obtained, which reveals the frequencies of the fault. Also, in order to evaluate the proposed method, we apply the used technique first on the simulated fault signal and then on the real signals in order to find the fault of the bearing components. In the continuation of the discussion, the experimental mode analysis method, which is one of the most powerful methods of vibration signal analysis, has been used in order to compare with the proposed method in identifying defects and analyzing vibration signals. The obtained results show that the external, internal rim, rolling element and bearing are healthy. The proposed method is superior to the experimental mode analysis in the detection of bearing defects.