No abstract

Bearings are an important part of rotating systems, and the long-term safe operation of mechanical equipment utilizing bearings is closely related to the bearing state; so, bearing fault diagnosis is of great significance. In this paper, a bearing fault diagnosis method based on comprehensive information divergence and improved BP (back propagation)-AdaBoost algorithm is proposed. First, the time domain, frequency domain, and time-frequency domain features under different states of the bearing are extracted to form a feature set. Then, the importance of each feature in fault classification is obtained by using the comprehensive information divergence index, and the feature sequence with decreasing importance is obtained. Finally, the most important features are selected as the input of the improved BP-AdaBoost classification model to train and obtain the bearing fault classification model. The experimental results show that the method has a good identification effect on bearing faults, and the stability of the model is high.

Milling, as a common machining method, is widely used in rough machining and final finishing of various materials. In this paper, according to the milling temperature produced in the milling process, the formula of heat distribution coefficient for workpiece milling is established. By means of Deform-3D finite element software to carry out orthogonal cutting simulation of workpiece, the influence of different machining parameters on milling heat distribution coefficient is studied, the optimal machining parameters are determined, and the milling temperature experiment is carried out to verify the simulation temperature. The experimental results show that the simulation temperature is very close to the experimental workpiece temperature, and the error is very small, which verifies the accuracy of the method. At the same time, the influence of different initial temperature of workpiece on the milling force and stability is also studied. The results show that proper heating of the workpiece can effectively improve the milling stability of the thin-walled parts.

Chatter in thin-walled parts is easy to occur in the process of machining, so the analysis of the stability of thin-walled parts has always been a research hotspot. In this paper, considering the influence of cutter eccentricity on milling force first, the coefficients of milling force were able to be identified by combining the milling force model with genetic algorithm. The results show that this method can obtain the milling force coefficients only by one experiment, and the accuracy is higher. Then the tool point Frequency Response Function (FRF) for a given combination can be calculated by using the Receptance coupling substructure analysis (RCSA) method that uses Timoshenko beam theory. Finally, the milling system can be divided into three types by aspect ratio. That is, when aspect ratio is less than 0.03, the system is considered to be a rigid tool-flexible workpiece system, but aspect ratio is between 0.03 and 0.2, the system is considered to be a flexible tool-flexible system, then aspect ratio is greater than 0.2, the system is considered to be a flexible cutter-rigid workpiece system.

The rotor system during its operation is susceptible to various faults such as unbalance, rub-impact, crack, and misalignment, which usually induce the rotor system to exhibit nonlinear behavior. Some linear diagnosis methods are unable to extract nonlinear characteristics of the faulty rotor system. However, existing nonlinear fault diagnosis methods can describe the nonlinear characteristics but cannot quantitatively indicate the severity of rub-impact faults. To address this issue, this study combines the nonlinear output frequency response functions weighted contribution rate (WNOFRFs) and JS divergence to develop an improved fault diagnosis approach, WNOFRFs based on the JS divergence (WNOFRFs-JS). And a superior NOFRFs-associated index JSRm is developed to indicate the severity of faults. In addition, a sensitive factor is defined to evaluate the sensitivity of the index. The performance of this approach is verified by an established dynamic model and a rotor rub-impact experimental rig. The results prove the effectiveness and merits of this approach for the identification of rotor rub-impact. JSRm is especially sensitive to rub-impact and can also quantitatively detect the severity of faults. The present approach can accurately and quantitatively identify the rub-impact rotor system. These advantages enable the improved WNOFRFs to be applied in the fault diagnosis and condition monitoring of rotating machinery and even other nonlinear systems.