The present invention will be further illustrated by specific examples below. Such as figure 1 As shown, this embodiment includes three parts: sending, receiving, and upper computer analysis and processing. Three test points are respectively installed with acceleration sensor 1; channel selection 7 is composed of CD4514 strobing and driving relay after ULN2003A power amplification; conditioning circuit 8 (such as image 3 (Shown) is composed of a charge amplifier 14, a preamplifier 15, a band pass filter 16, a post amplifier 17, an envelope detector 18, and a low pass filter 19. The charge amplifier 14 operation amplifier selects OP07; the preamplifier 15 and the post amplifier 16 select instrument amplifier AD620; the band pass filter 16 is a second-order band pass active filter with a center frequency of 5kHz and a bandwidth of 500Hz; the envelope detector 18 is composed of The low-pass filter 19 is composed of a Sallen-Key second-order low-pass filter with a cut-off frequency of 10 kHz. It is composed of a pass filter, a detector, a proportional addition and envelope circuit, and a voltage follower. The amplification gain is 1. The first single chip 9 and the second single chip 12 use MSP430F149, which is powered by 3.3V and has a built-in A/D converter, which can realize ultra-low power operation. CC1101 is used for wireless transmission 10 and wireless receiving 11, and the communication frequency is 433MHz. The data transmission with the host computer 13 selects MAX3232.
 Combined as figure 2 As shown, the vibration signal of the acceleration sensor 1 strobed by channel selection 7 is sampled by the first single-chip microcomputer 9 through the conditioning circuit 8 and the first single-chip 9 uses the SPI (Serial Peripheral Interface, serial peripheral interface) interface to perform A/D sampling. The digital signal is transmitted to the wireless transmitter 10, and the wireless receiver 11 receives the vibration signal, which is transmitted to the second single-chip 12 through the SPI interface, and then is connected to the upper computer 13 by the MAX3232.
 Continue as figure 1 As shown, the upper computer 13 obtains the rolling bearing vibration signal from the acceleration sensor 1, and obtains the maximum vibration value, mean square value, crest factor, pulse factor, form factor, kurtosis, and vibration waveform and frequency spectrum through the feature parameter extraction 3. Network analysis 4, failure pattern recognition 5 methods carry out data fusion and pattern recognition on characteristic variables, obtain fault information, and finally perform fault information processing 6 to realize fault diagnosis and fault type judgment.
 Neural network analysis 4 uses the BP neural network model, which is composed of three layers: the first layer is the input layer, which is composed of signal source nodes; the second layer is the hidden layer, the number of hidden units is determined by the problem described, the hidden unit The transformation function of is a non-negative nonlinear function that is radially symmetric and attenuated to the center point. The third layer is the output layer, which responds to the input mode. The actual input of the neural network analysis 4 is 6, which is composed of the root mean square value, The crest factor, form factor, kurtosis factor, margin factor, and impulse factor constitute the system feature vector. The hidden layer neurons are selected as 10, which are obtained from experience. For simplicity, a dual-value output network is used, and the output layer neurons are set to Two, the corresponding status codes are normal (0, 0) and fault (0, 1).
 Combine Figure 4 As shown, the known state vibration data 21 obtains training samples 23 through feature parameter extraction 22, which trains and deploys network weights together with target samples 20, and gradually approximates the expected output through constant changes in weights; the application process uses the deployed weights The value network classifies the actual test data to obtain the network weight 24. The real-time data 25 obtained from the wireless receiver 11 is obtained by the feature parameter extraction 26 to form the sample 27 to be identified. The identification 28 determines whether the rolling bearing is normal 29 or malfunctioning 30.
 The normalized data is trained using BP neural network. The data is used as a training sample as an input vector. After the input layer, intermediate layer and output layer of the network, the output result is compared with the target. If the error is If the value is within the allowable range of the maximum error, it means that the training has achieved the expected effect, and the training is ended. If the error is not within the allowable range of the maximum error, the data will be transferred to the input layer through the feedback of the BP neural network and a new round will start Of the loop. Until the requirements are met, a set of weight vector values can be obtained through the training of this neural network. Save the set of weight vector values in the designated file for the following pattern recognition.
 Combine Figure 5 The fault pattern recognition 5 is mainly composed of a knowledge acquisition module 32, a knowledge base module 33, a reasoning module 34, and an interpretation module 35. The input is the information source 31, the output is the diagnosis result 36, and the knowledge acquisition module 32 is used to analyze and organize Waveform data and extract its characteristic parameters to make it a form acceptable to the pattern recognition system. The knowledge base module 33 stores typical fault waveform characteristics and expert diagnosis rules. Specifically, it is a set of fault waveforms formed by modeling, and the reasoning module 34 uses The knowledge of the expert knowledge base analyzes and judges the measured waveform. The process of reasoning is the process of pattern recognition. The interpretation module 35 makes an appropriate analysis of the result of the reasoning and gives a final diagnosis. The characteristic parameters of a set of three-dimensional bearings to be identified are processed as the input vector of the network, and compared with the typical fault waveform feature, namely the target vector, which is close to within a certain error range, the state of the bearing to be identified is determined Which one belongs to, so as to distinguish the fault condition of the bearing.
 The wireless vibration monitoring system of the rolling machine rolling bearing of the present invention has the following characteristics:
 (1) Through in-depth research on the failure mechanism of rolling bearings, the detection points are determined according to the existing bearing damage and judgment mechanism. Three acceleration sensors are respectively arranged above the left bearing seat, lateral and right bearing seat, from the three Obtain the original bearing vibration information from different angles and levels. This information is selected by the channel and transformed by charge amplification, pre-amplification, band-pass filtering, post-amplification, envelope detection, low-pass filtering, A/D conversion, etc. Obtain the digital signal of vibration;
 (2) The digital signal is transmitted to the host computer by wireless transmission, and the sampled data is analyzed in the time domain and frequency domain to obtain the absolute maximum amplitude, root mean square value, crest factor, form factor, and kurtosis factor of the vibration , Margin factor and impulse factor; the root mean square value can reflect the total amount of vibration, is effective for the overall degradation of the bearing, has good stability, but is not sensitive to early fault signals; the crest factor can better reflect surface damage faults , Especially for the surface peeling at the initial stage; the form factor can reflect the increase of the frequency component; the kurtosis factor, the margin factor and the impulse factor are more sensitive to shock faults, especially when the fault occurs early, they have a significant increase; but After rising to a certain level, with the gradual development of faults, they will decrease instead. They have higher sensitivity to early faults, but the stability is not good; according to the different characteristics of these parameters, the sensitivity and stability of the system are taken into account. Root mean square value, crest factor, form factor, kurtosis factor, margin factor and impulse factor constitute the system feature vector;
 (3) The vibration information extracted from three different angles and levels is integrated to form a three-dimensional feature vector. Take this multi-dimensional feature vector as input, use a three-layer neural network as a diagnosis model, and diagnose through learning, training, and recognition. If the output is two, it is judged whether the rolling bearing is faulty;
 (4) The three-dimensional feature vector obtained by extracting the feature value after sampling the three sampling points is used as the input vector of the network. Through pattern recognition, the weight vector value obtained by neural network training is used for network calculation, and the vector value obtained in the output layer is compared with Compared with the target vector, which is close to within a certain error range, it is determined that the state of the bearing to be identified belongs to which kind, so as to identify the fault condition of the bearing;
 (5) In combination with the actual damage and failure of the bearing, conduct a comprehensive analysis and research on the historical record of bearing vibration and the corresponding failure mode, and summarize the peculiar law of the damage of the rolling machine rolling bearing and its relationship with the vibration characteristic parameters, etc. Improve the accuracy of bearing damage and fault diagnosis.
 In summary, it can be understood that the beneficial effect of the present invention is that the three-point acceleration detection obtains three different angles and levels of vibration information extracted to form a three-dimensional feature vector. The acceleration signal is effectively reduced by the channel selection module to reduce the complexity of the system and reduce the system Cost; Through wireless data transmission, the complicated wiring problems of the equipment are reduced, the anti-interference performance in the data transmission process is effectively improved, and the reliability of the system is improved; through neural network analysis and pattern recognition, combined with bearing vibration history and corresponding The failure mode of the bearing is comprehensively judged, the system misjudgment is reduced, and the accuracy of the fault diagnosis system is improved. At the same time, the system achieves reliable operation in ultra-low power consumption mode, prolongs battery life and greatly reduces equipment maintenance.
 The above are the best embodiments of the present invention. Based on the disclosure of the present invention, those of ordinary skill in the art can obviously think of some similarities and alternatives, which should fall within the protection scope of the present invention.