Health status assessment method of industrial robot reducer based on multi-source information fusion

Abstract

The present invention relates to an industrial robot reducer health status assessment method based on multi-source information fusion, which belongs to the technical field of industrial robots. First, various types of signals in the operation of the reducer are collected by different types of sensors, and then the time-frequency features of various signals are extracted. The time-frequency feature sets are used as training samples for BP neural networks and SVMs, and the feature sets of various signals are diagnosed respectively to obtain the recognition rate of each sensor for each type of fault. The D-S evidence theory is used to perform multi-source signal fusion on the local diagnosis results of various types of signals, and the diagnosis results based on the fusion of multi-source signals BP neural networks and SVM are obtained respectively. Finally, based on the classification results of multi-source signal fusion obtained by BP neural networks and SVM respectively, the D-S evidence theory is used to make decisions on the global fault diagnosis results obtained by BP neural networks and SVM to obtain the final diagnosis results. The present invention is based on the fusion of D-S decision layers, so that the accuracy and credibility of the diagnosis results are significantly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial robots, and in particular relates to a method for evaluating the health status of an industrial robot reducer based on multi-source information fusion. Background Art

[0002] Industrial robots are widely used in the manufacturing industry because they are stable, reliable and efficient. As the core component of the joints of industrial robots, the reducer ensures the precise movement of the industrial robot with its high-precision operation, and plays a vital role in the stable and reliable operation of the industrial robot. Once the reducer fails or its performance degrades, the precise movement of the industrial robot cannot be guaranteed, and the production workshop cannot operate normally, thus hindering the production of the enterprise and causing huge and serious losses to the enterprise. Therefore, it is very necessary to identify the health status of the reducer.

[0003] At present, most of the fault diagnosis technologies based on reducers are based on the identification of health status based on a single information source signal. This single sensor information processing technology is often affected by complex and changeable external factors in actual operation, and has great uncertainty or even errors. Therefore, it is difficult to meet the requirements of rapidity and accuracy of rotating machinery fault diagnosis by relying solely on a single information source.

[0004] Therefore, at this stage, it is necessary to design an industrial robot reducer health status assessment method based on multi-source information fusion to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide an industrial robot reducer health status assessment method based on multi-source information fusion, which is used to solve the technical problems existing in the above-mentioned prior art. At present, the fault diagnosis technology based on the reducer is mostly based on a single information source signal to identify the health status. This single sensor information processing technology is often affected by complex and changeable external factors during actual operation, and has great uncertainty or even errors. Therefore, it is difficult to meet the requirements of rapidity and accuracy of rotating machinery fault diagnosis by relying only on a single information source.

[0006] To achieve the above object, the technical solution of the present invention is:

[0007] The method for assessing the health status of an industrial robot reducer based on multi-source information fusion includes the following steps:

[0008] S1. Collect the corresponding vibration, temperature and noise signals by installing vibration, temperature and noise sensors on the industrial robot reducer, and use the motor driver itself as a sensor to obtain the corresponding current, voltage and torque signals of the reducer;

[0009] S2, calculating the time-frequency characteristic parameters of each collected signal;

[0010] S3, taking the characteristic parameters of the signal as input and the health status of the reducer as output to perform BP neural network learning and SVM classification, and dividing the health status of the reducer into four states: healthy state, gear fault, bearing fault, and shaft fault;

[0011] S4, using the redundant and complementary information of heterogeneous signals; testing the fault model based on BP neural network for the data of each signal in different health states, and obtaining the diagnosis rate in different health states, i.e., BPA value;

[0012] S5. Use the data of each signal under different health states to test the fault model based on SVM, obtain the classification accuracy under different health states, and use it as the BPA value to construct the corresponding evidence table as above, and calculate the fused BPA value;

[0013] S6. The obtained BP neural network classification results and SVM classification results are integrated through DS evidence theory to obtain the final diagnosis result.

[0014] Furthermore, in step S2, the time-frequency characteristic parameters include mean value, root mean square value, peak value, kurtosis index, margin index, pulse index, waveform index, average frequency, center frequency, and root mean square frequency.

[0015] Furthermore, in step S3, the BP neural network is trained by using the data of four different health states measured by each signal to establish a fault model; the Sigmoid function is selected as the output layer; and the classification result of the SVM is directly used as the BPA value in subsequent calculations based on the DS evidence theory.

[0016] Further, in step S4, 200 sets of data under different health states of each signal are used to test the fault model based on the BP neural network to obtain the diagnosis rate under different health states;

[0017] Different types of signals correspond to different bodies of evidence. DS evidence propositions are constructed with different types of signals, and the obtained diagnostic rate is used as the BPA value to construct a DS evidence table. E1 to E6 are set to represent six different types of evidence: vibration, temperature, noise, voltage, current, and torque. Then, the traditional DS evidence theory is used to perform multi-evidence fusion diagnosis. The calculation rules of the fused BPA value are as follows:

[0018]

[0019]

[0020] Where: m1(A1), m2(A2,)…, m6(A6) represent trust functions, which indicate the degree of trust in the proposition; A1, A2,…, A6 represent the set of evidences containing fault A; K represents the conflict factor, which reflects the conflict size between the evidences.

[0021] Furthermore, in step S6, the multi-source signal fusion result obtained by the BP neural network is used as new evidence E7, and the multi-source signal fusion result obtained by the SVM is used as new evidence E8, and then the two are fused through the DS evidence theory to obtain the final diagnosis result.

[0022] An industrial robot reducer health status assessment device based on multi-source information fusion, wherein the industrial robot reducer health status assessment device adopts the industrial robot reducer health status assessment method based on multi-source information fusion as described above to perform industrial robot reducer health status assessment.

[0023] A storage medium stores a computer program, which, when executed, executes the above-mentioned method for assessing the health status of an industrial robot reducer based on multi-source information fusion.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] One of the beneficial effects of this solution is that the present invention evaluates the health status of the reducer through the fault characteristics of multi-source data, uses a multi-decision model to perform information fusion through DS evidence theory decision-making, and deeply fuses the decision results through DS evidence theory decision-making to obtain the final diagnosis result. The reducer health status assessment method based on multi-source information fusion obtained through the above steps can significantly improve the accuracy of fault diagnosis of a single information source or a single decision model, has a high accuracy and credibility, and can meet the requirements for high accuracy of reducer status assessment in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Schematic diagram of the health status assessment method of industrial robot reducer based on multi-source information fusion.

[0027] Figure 2 This is a flow chart of the health status assessment method for industrial robot reducers based on multi-source information fusion. DETAILED DESCRIPTION

[0028] Below in conjunction with the appended Figure 1 -Attached Figure 2, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Example:

[0030] At present, most of the fault diagnosis technologies based on reducers are based on the identification of health status based on a single information source signal. This single sensor information processing technology is often affected by complex and changeable external factors in actual operation, and has great uncertainty or even errors. Therefore, it is difficult to meet the requirements of rapidity and accuracy of rotating machinery fault diagnosis by relying solely on a single information source.

[0031] like Figure 1 As shown in the figure, a method for assessing the health status of an industrial robot reducer based on multi-source information fusion is proposed. Compared with the diagnosis method based on a single information source, the proposed method can effectively improve the accuracy and credibility of the results and realize the health status detection and assessment of the industrial robot reducer. The schematic diagram and flow chart of the proposed method are shown in the figure. Figure 1 and Figure 2 shown.

[0032] The steps of the method are as follows:

[0033] Step 1: Collect the corresponding vibration, temperature and noise signals by installing vibration, temperature and noise sensors on the industrial robot reducer, and use the motor driver itself as a sensor to obtain the corresponding current, voltage and torque signals of the reducer.

[0034] Step 2: Calculate the time-frequency characteristic parameters of each signal collected. When processing fault signals, dimensional indicators such as mean, effective value, waveform factor, kurtosis, and skewness are often used. In order to monitor the operating status of diagnostic equipment, various dimensionless indicators are also widely used, mainly including: kurtosis index, margin index, waveform index, peak index, pulse index, etc. Dimensional indicators are relatively greatly affected by the working conditions of the equipment, and have high requirements for the constancy of the sensor measurement points; dimensionless indicators are less affected by external loads, measurement points and other factors, and are not sensitive to changes in factors such as the frequency and amplitude of the fault signal, so they are relatively adaptable. In diagnostic experiments, in order to ensure the comprehensiveness of information, dimensional indicators and dimensionless indicators are usually combined for consideration. Frequency domain indicators include average frequency, center frequency, root mean square frequency and other indicators. The operating status of rotating machinery can be evaluated based on the composition and size of these frequency components. Therefore, mean value, root mean square value, peak value, kurtosis index, margin index, pulse index, waveform index, average frequency, center frequency and root mean square frequency are selected as the ten time-frequency characteristic parameters of the signal. The calculation formulas of each index are shown in Table 1.

[0035] Table 1. Calculation formulas for signal characteristic parameters

[0036]

[0037]

[0038] In Table 1, x(n) is the time domain signal (such as acoustic emission, acceleration, velocity signal, etc.), f(n) is the frequency domain signal, and p n (f) represents the power spectrum of the signal, n=1,2,…,N, N is the number of sample points.

[0039] Step 3: Take the characteristic parameters of the signal as input and the health status of the reducer as output for BP neural network learning and SVM classification, and divide the health status of the reducer into four states: healthy state, gear fault, bearing fault, and shaft fault. The BP neural network is trained with the data of the above four different health states measured by each signal to establish a fault model. Since the output of the BP neural network is the basic probability distribution of the DS evidence theory, that is, the BPA value, the output of the network is required to be between [0, 1], so the output layer uses the Sigmoid function. The classification result of the SVM is directly used as the BPA value for subsequent calculations based on the DS evidence theory.

[0040] Step 4: Make full use of the redundant and complementary information of heterogeneous signals to further reduce decision uncertainty and improve the accuracy of fault diagnosis. Use 200 sets of data of each signal in different health states to test the fault model based on BP neural network and obtain the diagnosis rate under different health states. Different types of signals correspond to different bodies of evidence. DS evidence propositions are constructed with different types of signals, and the diagnosis rate obtained above is used as the BPA value to construct the DS evidence table. E1~E6 represent six different types of evidence: vibration, temperature, noise, voltage, current, and torque, respectively. Then, the traditional DS evidence theory is used to perform multi-evidence fusion diagnosis. The calculation rules of the fused BPA value are as follows:

[0041]

[0042]

[0043] Where: m1(A1), m2(A2), …, m6(A6) represent the trust function, which indicates the degree of trust in the proposition. A1, A2, …, A6 represent the set of evidences containing fault A, and K represents the conflict factor, which reflects the conflict between the evidences.

[0044] Step 5: Use 200 sets of data under different health states of each signal to test the SVM-based fault model, obtain the classification accuracy under different health states, and use it as the BPA value to construct the corresponding evidence table as above, and calculate the fused BPA value.

[0045] Step 6: In order to further improve the accuracy of the diagnosis results, the obtained BP neural network classification results and SVM classification results are fused through DS evidence theory. The multi-source signal fusion results obtained by the BP neural network are used as new evidence E7, and the multi-source signal fusion results obtained by the SVM are used as new evidence E8. The two are then fused through DS evidence theory to obtain the final diagnosis result.

[0046] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. An industrial robot reducer health status assessment method based on multi-source information fusion, characterized in that: The following steps are involved: S1. Collect the corresponding vibration, temperature and noise signals by installing vibration, temperature and noise sensors on the industrial robot reducer, and use the motor driver itself as a sensor to obtain the corresponding current, voltage and torque signals of the reducer; S2, calculating the time-frequency characteristic parameters of each collected signal; S3, taking the characteristic parameters of the signal as input and the health status of the reducer as output to perform BP neural network learning and SVM classification, and dividing the health status of the reducer into four states: healthy, gear fault, bearing fault, and shaft fault; S4, using the redundant and complementary information of heterogeneous signals; testing the fault model based on BP neural network for the data of each signal in different health states, and obtaining the diagnosis rate in different health states, i.e., BPA value; S5. Use the data of each signal under different health states to test the fault model based on SVM, obtain the classification accuracy under different health states, and use it as the BPA value to construct the corresponding evidence table as above, and calculate the fused BPA value; S6, the obtained BP neural network classification results and SVM classification results are integrated through DS evidence theory to obtain the final diagnosis result; In step S2, the time-frequency characteristic parameters include mean value, root mean square value, peak value, kurtosis index, margin index, pulse index, waveform index, average frequency, center frequency, and root mean square frequency; In step S3, the BP neural network is trained by the data of four different health states measured by each signal to establish a fault model; the Sigmoid function is selected as the output layer; the classification result of the SVM is directly used as the BPA value in the subsequent calculation based on the DS evidence theory; In step S4, 200 sets of data of each signal in different health states are used to test the fault model based on BP neural network to obtain the diagnosis rate in different health states; Different types of signals correspond to different bodies of evidence. DS evidence propositions are constructed with different types of signals, and the obtained diagnostic rate is used as the BPA value to construct a DS evidence table. E1 to E6 are set to represent six different types of evidence: vibration, temperature, noise, voltage, current, and torque. Then, the traditional DS evidence theory is used to perform multi-evidence fusion diagnosis. The calculation rules of the fused BPA value are as follows: Where: m1(A1), m2(A2,)…, m6(A6) represent trust functions, which indicate the degree of trust in the proposition; A1, A2,…, A6 represent the set of evidences containing fault A; K represents the conflict factor, which reflects the conflict size between the evidences.

2. The method for assessing the health status of an industrial robot reducer based on multi-source information fusion according to claim 1, characterized in that: In step S6, the multi-source signal fusion result obtained by the BP neural network is used as new evidence E7, and the multi-source signal fusion result obtained by the SVM is used as new evidence E8. The two are then fused through the DS evidence theory to obtain the final diagnosis result.

3. An industrial robot reducer health status assessment device based on multi-source information fusion, characterized in that: The industrial robot reducer health status assessment device adopts the industrial robot reducer health status assessment method based on multi-source information fusion as described in claim 1 or 2 to perform industrial robot reducer health status assessment.

4. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed, executes the method for assessing the health status of an industrial robot reducer based on multi-source information fusion as claimed in claim 1 or 2.

Images