Deep-forest-based mechanical bearing fault detection method

Abstract

The invention relates to a deep-forest-based mechanical bearing fault detection method, and belongs to the technical field of fault detection. The method comprises the following steps: firstly, performing a multi-granularity sampling strategy on acquired normal bearing data and various fault bearing data according to a vibration frequency f, grouping the data to obtain N groups of sampling data, performing feature extraction on the N groups of sampling data respectively to obtain an N groups of feature vectors Xs, labeling the N groups of feature vectors Xs respectively, integrating the N groups of feature vectors to obtain a feature vector X of the whole data, and finally, inputting the feature vector X into a deep forest to construct a cascaded structure to obtain a training model. The invention provides a multi-granularity grouping and feature extraction method based on the deep forest model, so that a training data set is effectively expanded; the phenomenon of loss of informationduring feature extraction is prevented; and the method is relatively high in applicability to mechanical bearing faults, and can be also applied to diagnosis of other types of mechanical faults.

Description

technical field [0001] The invention relates to a fault detection method for mechanical bearings based on deep forest, belonging to the technical field of fault detection. Background technique [0002] Bearings, gearboxes, etc. play an important role in industries that move and transmit torque transmitting machines, and they have applications in various fields such as aviation, aerospace, automotive, electric power, wind turbines, etc. As the core component of the machine, the bearing is easily damaged due to the long-term high-speed operation environment. Once a failure occurs, the loss of time and cost for the company will be huge. For complex systems, early detection of failure problems is critical, and taking remedial action to avoid dangerous situations can save valuable time and money. The integration between industrial manufacturing and the Internet is getting closer, and the diagnostic methods of mechanical faults play an important role in the reliability and safety...

AI Technical Summary

Problems solved by technology

There are many reasons for bearing failure, the collection of bearing signals is also diverse, and the collected fault signals are also uncertain.

If the feature extraction is directly performed on the original signal, it will bring about the lack of information other than the feature. For this problem, the traditional method will not only cause information loss and waste, but also cause the classifier to fail when the amount of data is not enough. Overfitting and other issues

Images