An equipment failure evaluation system and method based on multi-dimensional data of internet of things

Abstract

The application discloses a kind of equipment failure evaluation system based on multi-dimensional data of Internet of Things, and the multi-dimensional data such as mixed sound, axial vibration, radial vibration and bearing temperature in the working process of motor are collected and analyzed, to comprehensively evaluate motor failure.The application obtains the motor maintenance danger evaluation coefficient that multi-dimensional data shows by collecting the multi-dimensional data such as noise, vibration and temperature of motor and carrying out the analysis of multi-dimensional data, reflects the dangerous degree of continued use after motor failure, and realizes the quantitative display of dangerous degree, is convenient for comprehensive evaluation and prediction to motor failure, reaches the effect of early warning and maintenance.

Description

Technical Field

[0001] This invention belongs to the field of multi-dimensional data processing technology, and relates to a device fault assessment system and method based on multi-dimensional data from the Internet of Things. Background Technology

[0002] An electric motor, commonly known as a "motor," is an electromagnetic device that converts or transmits electrical energy based on the law of electromagnetic induction. The main function of an electric motor in a circuit is to generate driving torque, serving as a power source for electrical appliances or various machines. The main function of a generator in a circuit is to convert mechanical energy into electrical energy.

[0003] Currently, motor monitoring and management primarily focus on recording and storing operating parameters, lacking the ability to predict faults during operation. This results in poor accuracy and reliability in fault prediction. Furthermore, motors are susceptible to various faults that affect their lifespan. However, current fault diagnosis relies solely on manual experience and lacks experimental data, making it impossible to predict faults in advance. Consequently, accurate assessment of motor faults and prediction of the motor's continued operating lifespan based on current faults are insufficient, leading to inadequate motor monitoring and fault assessment, reduced motor lifespan, and hindering timely maintenance. Summary of the Invention

[0004] The purpose of this invention is to provide a device fault assessment system based on multi-dimensional data from the Internet of Things, which solves the problems in the prior art.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A device fault assessment system based on multi-dimensional data from the Internet of Things includes a noise source differentiation module, a vibration detection module, a bearing information detection module, a spectrum feature classification and processing module, a fault chain training interference module, a fault prediction and judgment module, and a multi-dimensional data assessment module.

[0007] The noise source differentiation module collects the mixed sound during the motor operation process in real time, and uses Fourier transform to perform spectrum analysis on the mixed sound during the motor operation process to obtain the spectrum characteristics of the separated bearing noise and the spectrum characteristics of the mechanical noise generated when the motor is working.

[0008] The vibration detection module uses an eddy current displacement sensor to collect the axial and radial vibration amplitudes of the motor in real time; the bearing information detection module uses a temperature sensor installed on the bearing to collect the temperature of the bearing surface.

[0009] The spectrum feature classification and processing module receives the spectrum features of the bearing noise and the mechanical noise generated when the motor is working after separation by the noise source differentiation module. The spectrum features of the bearing noise and the mechanical noise are compared with the spectrum features of the bearing noise and the mechanical noise at different fault levels, respectively, in order to filter out the bearing noise fault level and the mechanical noise fault level corresponding to the separated bearing noise.

[0010] The fault prediction and judgment module extracts the axial and radial vibration amplitudes of the motor, establishes a time-domain vibration signal graph based on the acquired axial and radial vibration amplitudes, and analyzes the time-domain vibration signal in the time-domain vibration signal graph using Fourier transform to obtain the frequency and phase of the axial vibration and the frequency and phase of the radial vibration. The basic parameters of the axial vibration and the basic parameters of the radial vibration are analyzed one by one to preliminarily predict whether the motor vibration is abnormal and to determine the degree of abnormality. The module also receives the temperature of the bearing surface, plots the temperature change curve of the bearing surface, and counts the maximum rate of temperature rise of the bearing surface, the cumulative duration of the bearing temperature being greater than the preset temperature W, and the bearing surface temperature after the bearing temperature is greater than the preset temperature W. The motor bearing seizure prediction coefficient is analyzed by the motor vibration abnormality judgment quantity and bearing temperature related parameter information.

[0011] The multi-dimensional data evaluation module extracts the bearing noise fault level and mechanical noise fault level screened by the spectrum feature classification processing module. Based on the bearing noise fault level and mechanical noise fault level, the bearing fault assessment coefficient corresponding to the bearing noise fault level and the mechanical fault assessment coefficient corresponding to the mechanical noise fault level are screened in sequence. The motor vibration abnormality degree, motor bearing seizure prediction coefficient and bearing surface temperature after the bearing temperature exceeds the preset temperature W are obtained by the fault prediction and judgment module. The multi-dimensional data evaluation model is used to predict the motor maintenance risk assessment coefficient for the current motor to continue to work.

[0012] Preferably, the method for determining the degree of abnormal motor vibration includes the following specific steps:

[0013] Step 1: Extract the amplitude, frequency, and phase of axial and radial vibrations;

[0014] Step 2: Determine whether the amplitude of the axial vibration is greater than k times the amplitude of the radial vibration. If it is greater than k times the amplitude of the radial vibration, mark the amplitude danger factor of the motor vibration as λ1. The preliminary value obtained from the experiment is 1.32. Otherwise, mark it as λ2. The preliminary value obtained from the experiment is 0.586.

[0015] Step 3: Analyze the ratio v between the axial vibration frequency f1 and the radial vibration frequency f2, and analyze the phase difference ψ between the axial vibration and the radial vibration, ψ=(2πf1T+w1)-(2πf2T+w2), where T is time, w1 and w2 are the initial phases of the axial vibration and the radial vibration, respectively.

[0016] Step 4: Combine the data from Step 2 and Step 3 to calculate the abnormal vibration coefficient. r is λ1 or λ2, v = f1 / f2.

[0017] Preferably, the motor bearing seizure prediction coefficient The calculation formula is as follows:

[0018] η1 represents the proportional coefficient for motor bearing seizure caused by vibration, η2 represents the proportional coefficient for motor bearing seizure caused by bearing temperature, η1 + η2 = 1, β1 represents the correlation interference coefficient for abnormal motor vibration caused by bearing temperature, β2 represents the correlation interference coefficient for abnormal motor vibration causing bearing temperature rise, and T represents the cumulative duration for which the bearing temperature exceeds the preset temperature W. 预 This represents the time interval during which the preset bearing temperature exceeds the preset temperature W. t1 and t2 represent the time points corresponding to the bearing temperature equaling the preset temperature W and the time points corresponding to the temperature exceeding the preset temperature W, respectively. t2 is greater than t1. max W' represents the maximum rate of temperature rise on the bearing surface, and W′ represents the bearing surface temperature after the bearing temperature exceeds the preset temperature W.

[0019] Preferably, the multi-dimensional data evaluation module is a comprehensive evaluation based on data processed from data collected by multiple sensors, and the multi-dimensional data evaluation model is as follows: a1, a2, a3, and a4 are the weighting coefficients for bearing noise, mechanical noise, motor seizure, and bearing temperature fault types, respectively. a1 + a2 + a3 + a4 = 1. X and Y are the bearing fault assessment coefficient and mechanical fault assessment coefficient, respectively. E is the abnormal motor vibration coefficient, n is 4, and T... 预 The maximum duration for which the preset bearing temperature exceeds the preset temperature W is set. max This is the maximum allowable temperature of the bearing surface. This represents the cumulative bearing surface temperature over time after the bearing temperature exceeds the preset temperature W. Let be the correlation interference coefficient between the j-th fault type and the ai-th fault type, where i = 1, 2, 3, 4, i.e., a1, a2, a3, a4 are bearing noise, mechanical noise, motor seizure, and abnormal bearing temperature, respectively. When i = j, It equals 0.

[0020] Preferably, the equipment fault assessment system further includes a fault chain training interference module. This module acquires the frequency of occurrence of each fault type in the motor fault type set A{a1,a2,...,ai,...,am} under the training duration, performs normalization analysis on the frequency of occurrence of each fault type, and analyzes the weight of each fault type. And based on the order in which each type of fault occurs, the number of interference effects C between each type of fault is counted. ai→aj To statistically determine the correlation interference coefficients among different fault types. X ai This represents the number of times each AI fault type occurs during the training period.

[0021] Preferably, the fault chain training interference module uses a clustering analysis method to cluster the number of interference effects between each fault type, and analyzes the correlation interference coefficient between each fault type, specifically including the following steps;

[0022] S1. After simulating training K times for each fault type, the weight of each fault type in this training process is calculated. The weight of each fault type is equal to the ratio of the number of times the fault type appears in the K training times to the number of training times K.

[0023] S2. Initially select Z types of faults as cluster centers;

[0024] S3. Establish the objective function Z represents the number of cluster centers, and Z represents the number of fault types. Let δ be the correlation interference degree between the fault type of the d-th sample and the g-th cluster center, and let p be the sum of the weights corresponding to all fault types in the simulation training in step S1. dg Let q be the distance between the d-th sample failure test and the g-th cluster center. d The weight of the fault type corresponding to the d-th sample fault test;

[0025] S4. Using the Lagrange multiplier method, derive the correlation interference influence matrix and the cluster center iteration formula for the objective function: D g R is the cluster center corresponding to the g-th fault type. g The weights corresponding to the types of faults in the training samples to be classified;

[0026] S5. Filter out the correlation interference coefficients between each fault type and the cluster center in the correlation interference influence matrix, and establish fault chains for each fault type with a correlation interference coefficient greater than 0.

[0027] Preferably, the equipment fault assessment system further includes a predictive damage tracking module. This module extracts the motor maintenance risk assessment coefficient obtained from the multi-dimensional data assessment module under the current motor operating state, and calculates the motor fault acceleration rate based on the motor maintenance risk assessment coefficient under the current motor state and the motor maintenance risk assessment coefficient at interval t3. And track and predict the lifespan of the motor if it continues to operate at the current motor failure surge rate. G t3 G is the motor sustaining risk assessment coefficient at time point t3. max The maximum allowable motor maintenance hazard assessment factor for the motor.

[0028] Preferably, a device fault assessment method based on multi-dimensional IoT data includes the following steps:

[0029] S1. Collect bearing temperature and mixed noise during motor operation, separate the mixed noise, and obtain bearing noise and mechanical noise;

[0030] S2. Screen the bearing noise and mechanical noise for fault levels to obtain the bearing noise fault level and mechanical noise fault level.

[0031] S3. Collect the axial vibration amplitude and radial vibration amplitude of the motor to establish a time-domain vibration signal diagram, and analyze the basic parameters of axial vibration and radial vibration through the time-domain vibration signal diagram;

[0032] S4. Extract the basic parameters of axial vibration and radial vibration from step S3 to determine the degree of motor vibration abnormality.

[0033] S5. Process the bearing surface temperature to obtain the maximum rate of temperature rise of the bearing surface, the cumulative duration of the bearing temperature being greater than the preset temperature W, and the bearing surface temperature after the bearing temperature is greater than the preset temperature W.

[0034] S6. Combine the data from steps S4 and S5 to analyze the motor's bearing seizure prediction coefficient, and use a multi-dimensional data evaluation model to determine the motor's maintenance risk assessment coefficient under the current motor continuing operation.

[0035] The beneficial effects of this invention are:

[0036] This application collects and analyzes multi-dimensional data such as motor noise, vibration, and temperature to obtain a motor maintenance risk assessment coefficient that displays the multi-dimensional data, reflecting the degree of danger of continuing to use the motor after a failure. It also quantifies the degree of danger, facilitating comprehensive assessment and prediction of motor failures, and achieving early warning and maintenance.

[0037] This application comprehensively processes motor vibration and bearing temperature to predict the possibility of motor seizure caused by these factors. It can predict and address motor seizure in advance, reducing the probability of seizure and thus improving the durability of the motor and preventing damage caused by seizure.

[0038] This application uses cluster analysis to analyze the various types of faults that cause motor failures, thereby obtaining the correlation interference coefficients between different types of faults. This provides a quantitative assessment of the degree of correlation between faults and thus provides a reliable basis for motor fault assessment.

[0039] This application obtains the motor maintenance risk assessment coefficient at two different time points, analyzes the motor failure surge rate based on the motor maintenance risk assessment coefficient at the two different time points, and then predicts the lifespan of the motor if it continues to work at the current motor failure surge rate. This enables the prediction of the lifespan of the motor under maintenance-free conditions, improves the accuracy of motor lifespan prediction, and promotes timely maintenance of motor failures. Detailed Implementation

[0040] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0041] Example 1

[0042] A device fault assessment system based on multi-dimensional data from the Internet of Things includes a noise source differentiation module, a vibration detection module, a bearing information detection module, a spectrum feature classification and processing module, a fault chain training interference module, a fault prediction and judgment module, and a multi-dimensional data assessment module.

[0043] The equipment referred to in this application refers to a motor. There are many types of motor failures. The main types of motor failures are bearing noise, mechanical noise, motor seizure, and abnormal bearing temperature. Therefore, this application focuses on analyzing the above four types of failures. By screening the above four types of failures, maintenance personnel can judge the motor failure situation and conduct fault assessment and prediction.

[0044] The noise source differentiation module collects the mixed sound during the motor operation process in real time, and uses Fourier transform to perform spectrum analysis on the mixed sound during the motor operation process to obtain the spectrum characteristics of the separated bearing noise and the spectrum characteristics of the mechanical noise generated during motor operation. The spectrum characteristics include amplitude, power and phase.

[0045] The vibration detection module uses an eddy current displacement sensor to collect the axial and radial vibration amplitudes of the motor in real time. The collected axial and radial vibration amplitudes are then sent to the fault prediction and judgment module.

[0046] The bearing information detection module uses a temperature sensor installed on the bearing to collect the temperature of the bearing surface and send the collected bearing surface temperature to the fault prediction and judgment module.

[0047] The spectrum feature classification processing module receives the spectrum features of the bearing noise and the mechanical noise generated by the motor after separation by the noise source differentiation module. It then compares the spectrum features of the bearing noise and the mechanical noise with the spectrum features of the bearing noise and the mechanical noise at different fault levels, respectively, to filter out the fault levels of the separated bearing noise and the mechanical noise. This process categorizes the fault levels corresponding to the sounds in the spectrum features, thus achieving qualitative processing of the fault levels.

[0048] The fault prediction and judgment module extracts the axial and radial vibration amplitudes of the motor. Based on the acquired axial and radial vibration amplitudes, it establishes a time-domain vibration signal graph and uses Fourier transform to analyze the time-domain vibration signal in the graph to obtain the frequency and phase of the axial and radial vibrations. It then performs a relative analysis of the basic parameters of the axial and radial vibrations one by one to preliminarily predict whether the motor vibration is abnormal and to determine the degree of abnormality. Additionally, it receives the bearing surface temperature, plots the temperature change curve, and statistically analyzes the maximum rate of temperature rise, the cumulative duration of the bearing temperature exceeding a preset temperature W, and the duration of the bearing temperature exceeding a preset temperature W. Let W be the bearing surface temperature. By analyzing the abnormality of motor vibration and bearing temperature-related parameters, a motor seizure prediction coefficient is derived. This coefficient reflects the likelihood of motor seizure under the combined influence of motor vibration and bearing surface temperature. Existing technologies only understand the relationship between motor seizure and motor vibration and bearing temperature, but cannot combine these factors for predictive assessment. A comprehensive prediction combining motor vibration and bearing temperature can more accurately analyze the likelihood of motor seizure, avoiding the limitations of existing technologies that rely solely on human experience. This allows for proactive motor intervention, reducing the likelihood of seizure and improving motor durability.

[0049] When a motor vibrates abnormally, it can indirectly cause the motor bearings to fail. Therefore, by quantifying the degree of abnormal motor vibration, we can intuitively and accurately grasp the motor fault.

[0050] The specific steps for determining the degree of abnormal motor vibration are as follows:

[0051] Step 1: Extract the amplitude, frequency, and phase of axial and radial vibrations;

[0052] Step 2: Determine whether the amplitude of the axial vibration is greater than k times the amplitude of the radial vibration. If it is greater than k times the amplitude of the radial vibration, mark the amplitude danger factor of the motor vibration as λ1. The preliminary value obtained from the experiment is 1.32. Otherwise, mark it as λ2. The preliminary value obtained from the experiment is 0.586.

[0053] Step 3: Analyze the ratio v between the axial vibration frequency f1 and the radial vibration frequency f2, and analyze the phase difference ψ between the axial vibration and the radial vibration, ψ=(2πf1T+w1)-(2πf2T+w2), where T is time, w1 and w2 are the initial phases of the axial vibration and the radial vibration, respectively.

[0054] Step 4: Combine the data from Step 2 and Step 3 to calculate the abnormal vibration coefficient. r is λ1 or λ2, v = f1 / f2. The abnormal vibration coefficient reflects the degree of abnormality of motor vibration. The larger the abnormal vibration coefficient, the greater the degree of abnormality of motor vibration, and the greater the possibility of bearing seizure caused by abnormal motor vibration.

[0055] Among them, the motor bearing seizure prediction coefficient The calculation formula is as follows:

[0056] η1 represents the proportional coefficient for motor bearing seizure caused by vibration, η2 represents the proportional coefficient for motor bearing seizure caused by bearing temperature, η1 + η2 = 1, β1 represents the correlation interference coefficient for abnormal motor vibration caused by bearing temperature, β2 represents the correlation interference coefficient for abnormal motor vibration causing bearing temperature rise, and T represents the cumulative duration for which the bearing temperature exceeds the preset temperature W. 预 This represents the time interval during which the preset bearing temperature exceeds the preset temperature W. t1 and t2 represent the time points corresponding to the bearing temperature equaling the preset temperature W and the time points corresponding to the temperature exceeding the preset temperature W, respectively. t2 is greater than t1. max W' represents the maximum rate of temperature rise on the bearing surface, and W′ represents the bearing surface temperature after the bearing temperature exceeds the preset temperature W.

[0057] The multi-dimensional data evaluation module extracts the bearing noise fault level and mechanical noise fault level screened by the spectral feature classification processing module. Based on the bearing noise fault level and mechanical noise fault level, it sequentially selects the bearing fault assessment coefficient corresponding to the bearing noise fault level and the mechanical fault assessment coefficient corresponding to the mechanical noise fault level. It also obtains the abnormal degree of motor vibration, the motor seizure prediction coefficient, and the bearing surface temperature after the bearing temperature exceeds the preset temperature W analyzed by the fault prediction and judgment module. The multi-dimensional data evaluation model predicts the current fault severity of the motor and obtains the motor maintenance risk assessment coefficient G, which reflects the degree of danger of continuing to use the motor after a fault, and quantifies the degree of danger. The multi-dimensional data evaluation module uses multi-dimensional detection data to comprehensively detect the motor, and achieves comprehensive data analysis by combining bearing temperature, mechanical noise, bearing noise, and motor seizure faults to realize a comprehensive assessment of motor faults. It can best show the degree of harm caused to the motor by continuing to operate under the current fault.

[0058] Among them, the bearing noise fault level and the bearing fault assessment coefficient are mutually mapped, and the mechanical noise fault level and the mechanical fault assessment coefficient are also mutually mapped. The bearing fault assessment coefficient and the mechanical fault assessment coefficient represent the probability of motor failure under the bearing noise fault level and the probability of motor failure under the mechanical noise level, respectively.

[0059] The multi-dimensional data evaluation module is a comprehensive evaluation based on data collected from multiple sensors. It achieves multi-dimensional data collection and analysis to improve the accuracy of motor fault assessment. The multi-dimensional data evaluation model is... a1, a2, a3, and a4 are the weighting coefficients for bearing noise, mechanical noise, motor seizure, and bearing temperature fault types, respectively. a1 + a2 + a3 + a4 = 1. X and Y are the bearing fault assessment coefficient and mechanical fault assessment coefficient, respectively. E is the abnormal motor vibration coefficient, n is 4, and T... 预 The maximum duration for which the preset bearing temperature exceeds the preset temperature W is set. max This is the maximum allowable temperature of the bearing surface. This represents the cumulative bearing surface temperature over time after the bearing temperature exceeds the preset temperature W. Let be the correlation interference coefficient between the j-th fault type and the ai-th fault type, where i = 1, 2, 3, 4, i.e., a1, a2, a3, a4 are bearing noise, mechanical noise, motor seizure, and abnormal bearing temperature, respectively. When i = j, It equals 0.

[0060] Example 2

[0061] Based on the order in which fault types appear, the degree of correlation between each fault type was initially analyzed, and a fault chain training interference module was designed.

[0062] The fault chain training interference module obtains the frequency of each fault type in the motor fault type set A{a1,a2,...,ai,...,am} under the training time. Normalization analysis is then performed on the frequency of each fault type to determine the weight of each fault type. And based on the order in which each type of fault occurs, the number of interference effects C between each type of fault is counted. ai→aj To statistically determine the correlation interference coefficients among different fault types. X ai This represents the number of times each AI fault type occurs during the training period.

[0063] Example 3

[0064] Since related faults have a causal relationship, when one fault occurs, another related fault will also occur. In order to improve the accuracy of motor fault assessment, the mean clustering algorithm is used to cluster the number of interference effects between each fault type to obtain the correlation interference coefficient between each fault type. The statistical correlation interference coefficient between each fault type is more accurate than that in Example 2, and eliminates human subjective factors.

[0065] For each fault type, K sample fault tests are simulated to facilitate the statistical analysis of the number of times fault type ai causes fault type aj in the motor when fault type ai occurs, thus obtaining the number of interference effects C between each fault type. ai→aj C ai →aj ≤K.

[0066] Among them, the fault chain training interference module uses clustering analysis to cluster the number of interference effects between each fault type, and analyzes the correlation interference coefficient between each fault type. Specifically, it includes the following steps.

[0067] S1. After simulating training K times for each fault type, the weight of each fault type in this training process is calculated. The weight of each fault type is equal to the ratio of the number of times the fault type appears in the K training times to the number of training times K.

[0068] S2. Initially select Z types of faults as cluster centers;

[0069] S3. Establish the objective function Z represents the number of cluster centers, and Z represents the number of fault types. Let δ be the correlation interference degree between the fault type of the d-th sample and the g-th cluster center, and let p be the sum of the weights corresponding to all fault types in the simulation training in step S1.dg Let q be the distance between the d-th sample failure test and the g-th cluster center. d The weight of the fault type corresponding to the d-th sample fault test;

[0070] S4. Using the Lagrange multiplier method, derive the correlation interference influence matrix and the cluster center iteration formula for the objective function: D g R is the cluster center corresponding to the g-th fault type. g The weights corresponding to the types of faults in the training samples to be classified;

[0071] S5. Filter out the correlation interference coefficients between each fault type and the cluster center in the correlation interference influence matrix, and establish fault chains for each fault type with a correlation interference coefficient greater than 0.

[0072] By using the mean clustering algorithm to cluster the number of interference effects among different fault types, the degree of correlation interference among different fault types can be accurately analyzed. This provides a quantitative representation of the degree of correlation in judging the mutual influence between faults, providing reliable correlation interference factors for motor fault assessment, and thus improving the accuracy of motor fault assessment.

[0073] Example 4

[0074] The predictive damage tracking module extracts the motor sustaining risk assessment coefficient obtained from the multi-dimensional data evaluation module under the current motor operating state. Based on the motor sustaining risk assessment coefficient under the current motor state and the motor sustaining risk assessment coefficient at interval t3, it calculates the motor fault excitation rate. And track and predict the lifespan of the motor if it continues to operate at the current motor failure surge rate. G t3 G is the motor sustaining risk assessment coefficient at time point t3. max The maximum allowable motor maintenance risk assessment coefficient is the coefficient for assessing the motor's current state. The higher the coefficient, the greater the likelihood of continued use of the motor.

[0075] Example 5

[0076] A device fault assessment method based on multi-dimensional IoT data, comprising the following steps:

[0077] S1. Collect bearing temperature and mixed noise during motor operation, separate the mixed noise, and obtain bearing noise and mechanical noise;

[0078] S2. Screen the bearing noise and mechanical noise for fault levels to obtain the bearing noise fault level and mechanical noise fault level.

[0079] S3. Collect the axial vibration amplitude and radial vibration amplitude of the motor to establish a time-domain vibration signal diagram, and analyze the basic parameters of axial vibration and radial vibration through the time-domain vibration signal diagram;

[0080] S4. Extract the basic parameters of axial vibration and radial vibration from step S3 to determine the degree of motor vibration abnormality.

[0081] S5. Process the bearing surface temperature to obtain the maximum rate of temperature rise of the bearing surface, the cumulative duration of the bearing temperature being greater than the preset temperature W, and the bearing surface temperature after the bearing temperature is greater than the preset temperature W.

[0082] S6. Combine the data from steps S4 and S5 to analyze the motor bearing seizure prediction coefficient, and use a multi-dimensional data evaluation model to determine the motor maintenance risk assessment coefficient under the current motor continuing to work, so as to reflect the harm caused to the motor by the motor continuing to work.

[0083] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in the claims, they should all fall within the protection scope of the present invention.

Claims

1. A device fault assessment system based on multi-dimensional data from the Internet of Things, characterized in that: It includes a noise source differentiation module, a vibration detection module, a bearing information detection module, a spectrum feature classification and processing module, a fault chain training interference module, a fault prediction and judgment module, and a multi-dimensional data evaluation module; The noise source differentiation module collects the mixed sound during the motor operation process in real time, and uses Fourier transform to perform spectrum analysis on the mixed sound during the motor operation process to obtain the spectrum characteristics of the separated bearing noise and the spectrum characteristics of the mechanical noise generated when the motor is working. The vibration detection module uses an eddy current displacement sensor to collect the axial and radial vibration amplitudes of the motor in real time; the bearing information detection module uses a temperature sensor, which is installed on the bearing to collect the temperature of the bearing surface. The spectrum feature classification and processing module receives the spectrum features of the bearing noise and the mechanical noise generated when the motor is working after separation by the noise source differentiation module. The spectrum features of the bearing noise and the mechanical noise are compared with the spectrum features of the bearing noise and the mechanical noise at different fault levels, respectively, in order to filter out the bearing noise fault level and the mechanical noise fault level corresponding to the separated bearing noise. The fault prediction and judgment module extracts the axial and radial vibration amplitudes of the motor, establishes a time-domain vibration signal graph based on the acquired axial and radial vibration amplitudes, and analyzes the time-domain vibration signal in the time-domain vibration signal graph using Fourier transform to obtain the frequency and phase of the axial vibration and the frequency and phase of the radial vibration. The basic parameters of the axial vibration and the basic parameters of the radial vibration are analyzed one by one to preliminarily predict whether the motor vibration is abnormal and to determine the degree of abnormality. The module also receives the temperature of the bearing surface, plots the temperature change curve of the bearing surface, and counts the maximum rate of temperature rise of the bearing surface, the cumulative duration of the bearing temperature being greater than the preset temperature W, and the bearing surface temperature after the bearing temperature is greater than the preset temperature W. The motor bearing seizure prediction coefficient is analyzed by the motor vibration abnormality judgment quantity and bearing temperature related parameter information. The multi-dimensional data evaluation module extracts the bearing noise fault level and mechanical noise fault level screened by the spectrum feature classification processing module. Based on the bearing noise fault level and mechanical noise fault level, the bearing fault assessment coefficient corresponding to the bearing noise fault level and the mechanical fault assessment coefficient corresponding to the mechanical noise fault level are screened in sequence. The abnormal degree of motor vibration, the motor bearing seizure prediction coefficient and the bearing surface temperature after the bearing temperature exceeds the preset temperature W are obtained by the fault prediction and judgment module. The multi-dimensional data evaluation model is used to predict the motor maintenance risk assessment coefficient for the current motor to continue to work. The motor bearing prediction coefficient The calculation formula is as follows: , This represents the proportional coefficient for motor bearing seizure caused by vibration. This represents the proportionality coefficient at which bearing temperature causes motor seizure. , This represents the correlation interference coefficient caused by abnormal motor vibration due to bearing temperature. This represents the correlation interference coefficient caused by abnormal motor vibration leading to an increase in bearing temperature. This represents the cumulative duration during which the bearing temperature is higher than the preset temperature W. This represents the time interval during which the preset bearing temperature exceeds the preset temperature W. t1 and t2 represent the time points corresponding to the bearing temperature equaling the preset temperature W and the time points corresponding to the temperature exceeding the preset temperature W, respectively. t2 is greater than t1. This represents the maximum rate of temperature rise on the bearing surface. This refers to the bearing surface temperature after the bearing temperature exceeds the preset temperature W. The multi-dimensional data evaluation module is a comprehensive evaluation based on data processed from data collected by multiple sensors. The multi-dimensional data evaluation model is as follows: a1, a2, a3, and a4 are the weighting coefficients corresponding to the fault types of bearing noise, mechanical noise, motor seizure, and bearing temperature, respectively, with a1 + a2 + a3 + a4 = 1. X and Y are the bearing fault assessment coefficient and the mechanical fault assessment coefficient, respectively, representing the probability of motor failure under the given bearing noise fault level and the probability of motor failure under the given mechanical noise level. E is the abnormal vibration coefficient of the motor, and n is 4. This refers to the maximum duration for which the preset bearing temperature exceeds the preset temperature W. This is the maximum allowable temperature of the bearing surface. This represents the cumulative bearing surface temperature over time after the bearing temperature exceeds the preset temperature W. Let be the correlation interference coefficient between the j-th fault type and the ai-th fault type, i=1,2,3,4, i.e., a1,a2,a3,a4 are bearing noise, mechanical noise, motor seizure, and abnormal bearing temperature, respectively. When i=j, It equals 0.

2. The device fault assessment system based on multi-dimensional IoT data according to claim 1, characterized in that: The method for determining the degree of abnormal motor vibration includes the following specific steps: Step 1: Extract the amplitude, frequency, and phase of axial and radial vibrations; Step 2: Determine if the amplitude of the axial vibration is greater than k times the amplitude of the radial vibration. If it is greater than k times the amplitude of the radial vibration, then mark the amplitude hazard factor of the motor vibration as [missing information]. The initial value obtained from the experiment was 1.32; otherwise, it was marked as... The preliminary value obtained from the experiment was 0.586; Step 3: Analyze the ratio v between the axial vibration frequency f1 and the radial vibration frequency f2, and analyze the phase difference between the axial vibration and the radial vibration. , T is time. and These are the initial phases of the axial vibration and the radial vibration, respectively. Step 4: Combine the data from Step 2 and Step 3 to calculate the abnormal vibration coefficient. r is or , .

3. The device fault assessment system based on multi-dimensional IoT data according to claim 1, characterized in that: The equipment fault assessment system also includes a fault chain training interference module, which acquires a set of motor fault types under a training duration. The frequency of each fault type was analyzed, and a normalization analysis was performed on the frequency of each fault type to determine the weight of each fault type. The number of interference effects between different fault types was counted based on the order in which they occurred. To statistically determine the correlation interference coefficients among different fault types. , This represents the number of times each AI fault type occurs during the training period.

4. The device fault assessment system based on multi-dimensional IoT data according to claim 1, characterized in that: The fault chain training interference module uses clustering analysis to cluster the number of interference effects between each fault type, and analyzes the correlation interference coefficient between each fault type. Specifically, it includes the following steps: S1. After simulating training K times for each fault type, the weight of each fault type in this training process is calculated. The weight of each fault type is equal to the ratio of the number of times the fault type appears in the K training times to the number of training times K. S2. Initially select Z types of faults as cluster centers; S3. Establish the objective function Z represents the number of cluster centers, and Z represents the number of fault types. Let g be the degree of interference between the fault type of the d-th sample and the g-th cluster center. The sum of weights corresponding to all fault types simulated in step S1. Let be the distance between the d-th sample failure test and the g-th cluster center. The weight of the fault type corresponding to the d-th sample fault test; S4. Using the Lagrange multiplier method, derive the correlation interference influence matrix and the cluster center iteration formula for the objective function: ; , The cluster center is the cluster center corresponding to the g-th fault type. The weights corresponding to the types of faults in the training samples to be classified; S5. Filter out the correlation interference coefficients between each fault type and the cluster center in the correlation interference influence matrix, and establish fault chains for each fault type with a correlation interference coefficient greater than 0.

5. A device fault assessment system based on multi-dimensional IoT data according to claim 1 or 4, characterized in that: The equipment fault assessment system also includes a damage prediction and tracking module. This module extracts the motor maintenance risk assessment coefficient obtained from the multi-dimensional data assessment module under the current motor operating state, and calculates the motor fault excitation rate based on the motor maintenance risk assessment coefficient under the current motor state and the motor maintenance risk assessment coefficient at interval t3. And track the predicted lifespan of the motor if it continues to operate at the current motor failure surge rate. , The motor sustaining risk assessment coefficient at time point t3. The maximum allowable motor maintenance hazard assessment factor for the motor.

6. A device fault assessment method based on multi-dimensional data from the Internet of Things, characterized in that: The specific steps are as follows: S1. Collect bearing temperature and mixed noise during motor operation, separate the mixed noise, and obtain bearing noise and mechanical noise; S2. Screen the bearing noise and mechanical noise for fault levels to obtain the bearing noise fault level and mechanical noise fault level. S3. Collect the axial vibration amplitude and radial vibration amplitude of the motor to establish a time-domain vibration signal diagram, and analyze the basic parameters of axial vibration and radial vibration through the time-domain vibration signal diagram; S4. Extract the basic parameters of axial vibration and radial vibration from step S3 to determine the degree of motor vibration abnormality. S5. Process the bearing surface temperature to obtain the maximum rate of temperature rise of the bearing surface, the cumulative duration of the bearing temperature being greater than the preset temperature W, and the bearing surface temperature after the bearing temperature is greater than the preset temperature W. S6. Combine the data from steps S4 and S5 to analyze the motor bearing seizure prediction coefficient, and use a multi-dimensional data evaluation model to determine the motor maintenance risk assessment coefficient under the current motor continuing to work. The motor bearing prediction coefficient The calculation formula is as follows: , This represents the proportional coefficient for motor bearing seizure caused by vibration. This represents the proportionality coefficient at which bearing temperature causes motor seizure. , This represents the correlation interference coefficient caused by abnormal motor vibration due to bearing temperature. This represents the correlation interference coefficient caused by abnormal motor vibration leading to an increase in bearing temperature. This represents the cumulative duration during which the bearing temperature is higher than the preset temperature W. This represents the time interval during which the preset bearing temperature exceeds the preset temperature W. t1 and t2 represent the time points corresponding to the bearing temperature equaling the preset temperature W and the time points corresponding to the temperature exceeding the preset temperature W, respectively. t2 is greater than t1. This represents the maximum rate of temperature rise on the bearing surface. This refers to the bearing surface temperature after the bearing temperature exceeds the preset temperature W. The multi-dimensional data evaluation module is a comprehensive evaluation based on data processed from data collected by multiple sensors. The multi-dimensional data evaluation model is as follows: a1, a2, a3, and a4 are the weighting coefficients corresponding to the fault types of bearing noise, mechanical noise, motor seizure, and bearing temperature, respectively, with a1 + a2 + a3 + a4 = 1. X and Y are the bearing fault assessment coefficient and the mechanical fault assessment coefficient, respectively, representing the probability of motor failure under the given bearing noise fault level and the probability of motor failure under the given mechanical noise level. E is the abnormal vibration coefficient of the motor, and n is 4. This refers to the maximum duration for which the preset bearing temperature exceeds the preset temperature W. This is the maximum allowable temperature of the bearing surface. This represents the cumulative bearing surface temperature over time after the bearing temperature exceeds the preset temperature W. Let be the correlation interference coefficient between the j-th fault type and the ai-th fault type, i=1,2,3,4, i.e., a1,a2,a3,a4 are bearing noise, mechanical noise, motor seizure, and abnormal bearing temperature, respectively. When i=j, It equals 0.

Images