Method and device for diagnosing health state of electromechanical equipment based on machine learning

Abstract

The invention relates to the technical field of electrical fault detection, in particular to an electromechanical equipment health state diagnosis method and device based on machine learning, and the method comprises the following steps: collecting the current, voltage and temperature parameters of equipment, calculating the real-time average value and standard deviation of each parameter through numerical processing, and obtaining a parameter monitoring result; and monitoring a result based on the parameters. The current, voltage and temperature parameters of the equipment are collected in real time, numerical processing is carried out on each parameter, the real-time average value and the standard deviation are calculated, and accurate basic data support is provided for state monitoring. Based on the monitoring data, parameter abnormity is judged through threshold analysis, abnormal changes can be dynamically recognized in the equipment operation process, and the timeliness and accuracy of fault discovery are improved. In combination with past performance data, the future change trend of parameters is predicted through a regression algorithm, early warning for equipment operation can be formed, and problem expansion or out-of-control is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical fault detection, and in particular to a diagnostic method and device for the health status of electromechanical equipment based on machine learning. Background Art

[0002] The technical field of electrical fault detection belongs to an important branch of power engineering and equipment maintenance, mainly involving the real-time monitoring, fault diagnosis and predictive analysis of the operating status of power systems and electrical equipment. By collecting operating parameters such as current, voltage, power, temperature, and vibration of electrical equipment, combined with technologies such as signal processing, pattern recognition, and machine learning, the fault point is accurately located, the fault type is judged, and the health status of the equipment is evaluated.

[0003] Although the existing technologies can monitor the operating status of electrical equipment in real time, they are mostly limited to collecting single or a small number of operating parameters of the equipment, lacking the dynamic analysis of the correlation relationships between multiple parameters, which may lead to an incomplete judgment of the scope of the fault impact. The fault diagnosis method is usually based on a simple comparison of static parameter thresholds and cannot cope with the complex dynamic change relationships between parameters during equipment operation, easily resulting in misjudgment or missed judgment. Summary of the Invention

[0004] The purpose of the present invention is to solve the disadvantages existing in the prior art, and to propose a diagnostic method and device for the health status of electromechanical equipment based on machine learning.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solution. A diagnostic method for the health status of electromechanical equipment based on machine learning includes the following steps:

[0006] Collect the current, voltage, and temperature parameters of the equipment, and through numerical processing, calculate the real-time average value and standard deviation of each parameter to obtain the parameter monitoring result; based on the parameter monitoring result, determine whether the parameter is abnormal and generate an abnormal parameter identification result;

[0007] Based on the abnormal parameter identification result, select parameters for analysis, combine past performance data, and use a regression algorithm to predict the change trend of the parameters within a future time period to generate a parameter trend prediction result; evaluate the parameter trend prediction result to judge the health status of the equipment and obtain an equipment health evaluation result;

[0008] Based on the equipment health evaluation result, identify the key influencing factors of the equipment performance, extract the key influencing factors to obtain a key influencing factor identification result; adjust the maintenance strategy according to the key influencing factor identification result, formulate targeted maintenance measures, and generate an optimized maintenance strategy result;

[0009] Based on the optimization result of the maintenance strategy, implement maintenance measures, monitor the maintenance effect through real-time data, calculate the performance improvement ratio of the equipment after maintenance, and obtain the maintenance effect evaluation result.

[0010] Preferably, the steps for obtaining the parameter monitoring result are as follows:

[0011] Collect the current, voltage, and temperature data of the equipment through sensors, record the current value, voltage value, and temperature value of each collection, and obtain the original collection data;

[0012] According to the original collection data, calculate the real-time average value and standard deviation of the current data, the real-time average value and standard deviation of the voltage data, and the real-time average value and standard deviation of the temperature data, and generate the real-time monitoring result;

[0013] Based on the real-time monitoring result, integrate the real-time monitoring results of the current data, the real-time monitoring results of the voltage data, and the real-time monitoring results of the temperature data to form the parameter monitoring result.

[0014] Preferably, the steps for obtaining the abnormal parameter identification result are as follows:

[0015] Based on the parameter monitoring result, intercept the current monitoring numerical sequence, voltage monitoring numerical sequence, and temperature monitoring numerical sequence within the most recent 30 days, divide them into fixed equal-width time periods by day respectively, extract the maximum value, minimum value, first and last difference, and the number of monotonic intervals of each time period, and generate the current fluctuation profile group, voltage fluctuation profile group, and temperature fluctuation profile group;

[0016] According to the current fluctuation profile group, voltage fluctuation profile group, and temperature fluctuation profile group, calculate the co-disturbance intensity value, and the calculation formula is:

[0017]

[0018] where D c is the co-disturbance intensity value of the c-th type of parameter, A cu is the maximum value of the u-th time period, B cu is the minimum value of the u-th time period, C cu is the first and last difference, Z cu is the number of consecutive segments of monotonic increase or decrease within the u-th time period, Y cu is the direction switching frequency between two adjacent peaks, and m is the total number of time periods counted;

[0019] According to the co-disturbance intensity value, set the stable value range for the current monitoring parameter, voltage monitoring parameter, and temperature monitoring parameter respectively, and judge whether the co-disturbance intensity value continues to be higher than the stable value range to generate the abnormal parameter identification result.

[0020] Preferably, the steps for obtaining the parameter trend prediction result are as follows:

[0021] Select the parameters showing abnormal trends from the abnormal parameter identification results, including abnormal current, abnormal voltage, and abnormal temperature data, to obtain a list of selected abnormal parameters;

[0022] Based on the list of selected abnormal parameters, summarize the historical change patterns and the context of abnormal occurrences to obtain an integrated historical performance dataset;

[0023] According to the integrated historical performance dataset, apply linear regression to predict the future change trends of the abnormal parameters, judge the change range and rate, and generate the parameter trend prediction result.

[0024] Preferably, the steps for obtaining the equipment health assessment result are as follows:

[0025] Based on the parameter trend prediction result, evaluate the health impact degree of each parameter, and calculate the equipment health score. The calculation formula is:

[0026]

[0027] where H represents the equipment health score, J represents the number of parameters, P j represents the predicted value of the j-th parameter, T j represents the target threshold of the j-th parameter, S j represents the sensitivity coefficient of the j-th parameter;

[0028] According to the equipment health score, judge whether the equipment health score is lower than the safety threshold. If it is lower, determine that the equipment status is a potential risk, and generate the equipment health assessment result.

[0029] Preferably, the steps for obtaining the key influencing factor identification result are as follows:

[0030] Based on the equipment health assessment result, calculate the factor scores of each index;

[0031] According to the factor scores, select the performance index with the highest score as the key influencing factor of the equipment performance to obtain the key influencing factor identification result.

[0032] Preferably, the steps for obtaining the maintenance strategy optimization result are as follows:

[0033] Based on the key influencing factor identification result, classify and organize the corresponding equipment components, operating environments, and operating parameters to obtain a list of key maintenance elements;

[0034] According to the list of key maintenance elements, analyze the maintenance priorities of each key maintenance element in each operating state, and match the maintenance strategies to form a maintenance task allocation plan;

[0035] Based on the maintenance task allocation scheme, formulate operation measures, design corresponding maintenance operations for each key maintenance element, and match the maintenance measures with the equipment operation plan to generate the optimized result of the maintenance strategy.

[0036] Preferably, the steps for obtaining the maintenance effect evaluation result are as follows:

[0037] According to the optimized result of the maintenance strategy, adjust, repair or replace the components of the equipment, record the operation data and the changes in the equipment status during each maintenance process, and generate a maintenance operation record.

[0038] Based on the maintenance operation record, evaluate the operation performance of the equipment after maintenance by monitoring the operation data of the equipment, including the change trends of current, voltage and temperature, and generate a performance evaluation report after maintenance.

[0039] Based on the performance evaluation report after maintenance, calculate the improvement ratio of the equipment performance compared with that before maintenance, analyze the change trend of the equipment performance under different operation states, and obtain the maintenance effect evaluation result.

[0040] The present invention provides a diagnostic device, including:

[0041] A data acquisition module, which collects the current, voltage and temperature data of the motor, calculates the real-time average value and standard deviation of the data, and generates a parameter monitoring result.

[0042] An anomaly analysis module, which based on the parameter monitoring result, compares each item of data with a preset threshold value, identifies the data beyond the normal range, and generates an abnormal parameter identification result.

[0043] A trend prediction module, which uses the abnormal parameter identification result, analyzes the historical data of related parameters, predicts the change trend within a future time period, and generates a parameter trend prediction result.

[0044] A health assessment module, which analyzes the parameter trend prediction result, judges the current health state of the equipment, and generates an equipment health assessment result.

[0045] A maintenance strategy optimization module, which based on the equipment health assessment result, analyzes the key influencing factors of the equipment performance, extracts and analyzes the key influencing factors, formulates or adjusts the maintenance measures of the equipment, and generates an optimized result of the maintenance strategy.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0047] The present invention collects the current, voltage and temperature parameters of the device in real time, processes the numerical values of each parameter, calculates the real-time average value and standard deviation, and provides accurate basic data support for condition monitoring. Based on these monitoring data, threshold analysis is used to determine parameter anomalies, which can dynamically identify abnormal changes during the operation of the device, improving the timeliness and accuracy of fault discovery. Combining past performance data, the future change trend of the parameters is predicted through a regression algorithm, which can form a warning for the operation of the device to avoid the expansion or out-of-control of problems. And by extracting the key influencing factors, the parameters affecting the device performance are effectively identified, helping to eliminate secondary interferences and optimizing the accuracy of the maintenance process. According to the key influencing factors, the maintenance strategy is adjusted, and targeted maintenance measures are designed, improving the scientificity and rationality of the strategy implementation. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0049] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0050] Please refer to Figure 1 , the present invention provides a technical solution, a diagnostic method for the health status of electromechanical equipment based on machine learning, including the following steps:

[0051] Collect the current, voltage and temperature parameters of the device, calculate the real-time average value and standard deviation of each parameter through numerical processing to obtain the parameter monitoring results; based on the parameter monitoring results, determine whether there are anomalies in the parameters and generate the abnormal parameter identification results.

[0052] Based on the abnormal parameter identification results, select parameters for analysis, combine past performance data, and use a regression algorithm to predict the change trend of the parameters within a future time period to generate the parameter trend prediction results; evaluate the parameter trend prediction results to judge the health status of the device and obtain the device health assessment results.

[0053] Based on the device health assessment results, identify the key influencing factors of the device performance, extract the key influencing factors to obtain the key influencing factor identification results; adjust the maintenance strategy according to the key influencing factor identification results, formulate targeted maintenance measures, and generate the maintenance strategy optimization results.

[0054] Based on the maintenance strategy optimization results, implement the maintenance measures, monitor the maintenance effect through real-time data, and calculate the performance improvement ratio of the device after maintenance to obtain the maintenance effect assessment results.

[0055] The steps to obtain the parameter monitoring results are as follows:

[0056] Collect the current, voltage, and temperature data of the device through sensors, record the current value, voltage value, and temperature value of each collection, and obtain the original collection data.

[0057] According to the original collection data, calculate the real-time average value and standard deviation of the current data, the real-time average value and standard deviation of the voltage data, and the real-time average value and standard deviation of the temperature data, and generate the real-time monitoring results.

[0058] Based on the real-time monitoring results, integrate the real-time monitoring results of the current data, the real-time monitoring results of the voltage data, and the real-time monitoring results of the temperature data to form the parameter monitoring results.

[0059] Specifically, based on a pre-determined layout plan, select multiple current sensors, voltage sensors, and temperature sensors and install them at key positions of the device respectively. Monitor the real-time values during the operation of the device and record the current value, voltage value, and temperature value at the current moment. Collect data several times per second and save these data as the original collection data. Define the value ranges of parameters such as current and voltage preliminarily. For example, compare the current value with the range of 0A to 5A, the voltage value with the range of 0V to 24V, and the temperature value with the range of 0°C to 90°C. These parameter ranges are selected based on past operation experience and industry specifications. If any monitored data falls into or exceeds these ranges during the collection process, mark it, and retain the corresponding record after marking. If the temperature value exceeds 90°C, manual secondary confirmation is required to determine whether it is due to actual collection in a high-temperature section or sensor abnormality. The temperature threshold here is obtained by combining the device's heat load design standard and empirical research. If manual reconfirmation shows that the temperature is normal, maintain the original marked state. If it is indeed a sensor failure, do not retain this record. All the remaining normally collected data information is uniformly stored in the system to form all the original collection data.

[0060] After obtaining all the original acquisition data, the numerical sum of the current data is calculated item by item and divided by the number of valid data to obtain the real-time average value of the current. The square sum of the difference between each current data and the real-time average value is divided by the number of valid data and then the square root is taken to obtain the standard deviation. The number of valid data here is the sum of the abnormal records caused by sensor failures excluded. The calculation method of the standard deviation is set according to the characteristics of the Gaussian distribution. If it is found during the calculation that the deviation of the real-time average value from the historical average value is too large, the deviation points in the original records will be verified again. Similar calculation methods are also used for the voltage data and temperature data, and their real-time average values and standard deviations are obtained respectively. Since the voltage range and temperature range have been set in the early stage, when a certain calculation result shows that the real-time average value exceeds the upper limit of the range, the judgment threshold will be adjusted according to the empirical formula. This empirical formula is obtained by performing linear regression analysis on the historical data collected multiple times during the actual operation process. If the value range deduced in the empirical formula is exceeded, it is necessary to further check whether there are unreasonable fluctuations in the marked information and relevant data. After confirmation of availability, the average values and standard deviations of these three parameters can be summarized as the real-time monitoring results.

[0061] According to the current average value and current standard deviation, voltage average value and voltage standard deviation, and temperature average value and temperature standard deviation in the real-time monitoring results, the corresponding values at the current moment are read in sequence and merged in the same table. The current part and the voltage part are aligned based on the time stamp as the association key, and then the temperature part is also associated according to the same time stamp. If the current information is missing at certain moments, blank placeholders are reserved and invalid identifiers are noted. These invalid identifiers can be analyzed again. The association method used here is to match based on the time tag of each acquisition record and the corresponding sensor source. When there are multiple records at the same moment, the value collected first is taken as the main value and the subsequent multiple repeated records are noted. If there is only a single record at the same moment, it is directly included in the integration scope. After all the records are checked, a parameter monitoring result containing comprehensive information of current, voltage, and temperature will be generated.

[0062] The steps to obtain the abnormal parameter identification result are as follows:

[0063] Based on the parameter monitoring results, intercept the current monitoring numerical sequence, voltage monitoring numerical sequence, and temperature monitoring numerical sequence within the most recent 30 days. Each sequence is divided into fixed equal-width time periods on a daily basis, and the maximum value, minimum value, first and last difference, and the number of monotonic intervals are extracted for each period to generate a current fluctuation profile group, a voltage fluctuation profile group, and a temperature fluctuation profile group.

[0064] According to the current fluctuation profile group, voltage fluctuation profile group, and temperature fluctuation profile group, calculate the co-disturbance intensity value. The calculation formula is:

[0065]

[0066] Among them, D c is the co-disturbance intensity value of the c-th type of parameter, A cu is the maximum value in the u-th time period, B cu is the minimum value in the u-th time period, C cu is the first and last difference, Z cu is the number of consecutive segments that are monotonically increasing or decreasing within the u-th time period, Y cu is the direction switching frequency between two adjacent peaks, and m is the total number of statistical time periods;

[0067] Based on the co-disturbance intensity value, set the stable value ranges for the current monitoring parameter, voltage monitoring parameter, and temperature monitoring parameter respectively, and determine whether the co-disturbance intensity value continuously exceeds the stable value range to generate the abnormal parameter identification result.

[0068] Specifically, based on the parameter monitoring results, intercept the current monitoring numerical sequence, voltage monitoring numerical sequence, and temperature monitoring numerical sequence within the most recent thirty days. Divide each day into several fixed equal-width time periods respectively, group the current values, voltage values, and temperature values at all corresponding moments within the day in the order of time periods, statistically analyze the numerical distribution within each group for each time period, successively select the maximum value and minimum value in each time period, and perform a comparison between adjacent records to obtain the number of monotonic intervals. The determination method of the number of monotonic intervals is to successively detect the increasing or decreasing trend of adjacent observation points within each time period. Any sequence that continuously maintains the same direction of change is regarded as a monotonic interval, and the number of such intervals is recorded. Denote the difference between the first observation point and the last observation point within the same time period as the first and last difference, and mark the above-mentioned maximum value, minimum value, first and last difference, and the number of monotonic intervals in the corresponding statistical entries respectively. The data recording volume for each time period is based on all valid observation points within that time period of the day. When there are invalid marks, they will be excluded before performing numerical analysis. Then, after the statistics are completed, organize the current parameter, voltage parameter, and temperature parameter within each time period respectively to generate the current time period characteristics, voltage time period characteristics, and temperature time period characteristics entries, and then arrange them in sequence according to different time period indexes and merge them into the monthly time period distribution situation, finally forming the basic set of three types of fluctuation information. Summarize the current part and identify its maximum and minimum difference, monotonic interval count, and first and last difference value. Perform the same operations on the voltage and temperature parts and align them with the current sequence. Store the obtained time period records in sequence and complete the result statistics of the daily time period distribution within all thirty days to generate the current fluctuation profile group, voltage fluctuation profile group, and temperature fluctuation profile group.

[0069] Formula: The advantage of the formula is that by introducing this ratio in the numerator part to quantify the relative fluctuation degree between the maximum value and the minimum value within each time period, and combining with Perform multi-dimensional analysis on the head-tail difference and direction switching frequency, thereby integrating the amplitude and direction conversion under the same numerical structure, enabling the unified measurement of the perturbation patterns of different parameters within a time period, and further helping to promptly determine whether the fluctuation intensity of a certain type of parameter is significant.

[0070] A cu The acquisition step of the parameter is as follows. For the u-th time period of a certain type of parameter c, select the maximum value from all the observed data within the time period as A cu , which is obtained by directly scanning the time period observation records. The number of data records is generally several samples per second to form an observation set of a certain duration, and A can be obtained through the maximum value operation. cu , for example, within the second time period of the current parameter, if this time period contains 300 sampling points and the maximum sampling point is 4.6A, then A 1,2 is recorded as 4.6A.

[0071] B cu The acquisition step of the parameter is as follows. For the same time period interval as A cu , select the minimum observed value within this time period as B cu , which is also obtained by scanning the observation records. B can be obtained by performing a minimum value extraction on all the sampling points once. cu , for example, within the second time period of the same current parameter, if the minimum sampling value is 3.9A, then B 1,2 is recorded as 3.9A.

[0072] C cu The acquisition step of the parameter is as follows. Read the difference between the first observed point and the last observed point within the same time period, and record it as C cu , this value can be obtained by indexing and reading the first and last records of all the sampling points within the time period and performing a subtraction operation to obtain the head-tail difference. If the first sampling value of a certain time period of the voltage parameter is 22.0V and the last sampling value is 22.3V, then C 2,u is recorded as 0.3V.

[0073] Z cu The acquisition step of the parameter is as follows. Count the number of consecutive segments that are monotonically increasing or decreasing within this time period. Traverse the sequence from the first observed point to the last observed point, and compare the size relationship of adjacent observed values in turn. Each time there is a switch from increasing to decreasing or from decreasing to increasing, mark the end of a new monotonic segment and start a new monotonic segment. Finally, record the total number of monotonic segments within this time period as Z cu , for example, within the seventh time period of the temperature parameter, there are 100 sampling points in total, and there are three inflection points from rising to falling during the inspection. Then at this time, Z 3,7 is 4 monotonic segments. The same operation needs to be performed on each time period at the monthly data level to obtain all Zcu 。

[0074] Y cu The steps for obtaining the parameter are as follows: obtain the direction switching frequency between two adjacent peaks within this period by scanning the wave peaks in each period to determine the number of wave peaks, then count the frequency of upward or downward reversals between wave peaks, and divide this frequency by the total number of samples in the current period to generate a numerical quantization result of direction switching. If there are a total of 5 wave peaks in the voltage period and 3 of them change from rising peaks to falling peaks, the total number of direction switches is 3. Divide it by the number of samples of this parameter in the current period to obtain Y cu , for example, if the number of sampling points in the current period is 300 and the total number of direction switches is 3, then Y cu = 3 / 300 = 0.01.

[0075] The steps for obtaining the m parameter are as follows: count the total number of all periods corresponding to a certain type of parameter within thirty days and denote it as m. This total number can determine the period division method according to the monitoring strategy. For example, if each hour is taken as a period, there are a total of 720 periods in a month; if every two hours is taken as a period, 360 periods are obtained.

[0076] Calculation process:

[0077] Let c = 1 correspond to the current parameter, and select m = 3 periods for the calculation example. For example, the obtained period maximum values A 1,1 = 4.8A, A 1,2 = 4.5A, A 1,3 = 4.2A, the minimum value B 1,1 = 3.9A, B 1,2 = 3.6A, B 1,3 = 3.8A, the first and last difference C 1,1 = 0.2A, C 1,2 = -0.3A, C 1,3 = -0.1A, the number of monotonic segments Z 1,1 = 2, Z 1,2 = 1, Z 1,3 = 2, the direction switching frequency Y 1,1 = 0.01, Y 1,2 = 0.02, Y 1,3 = 0.01. Substitute these values into the formula item by item. The summation process of the numerator part is as follows:

[0078] The first period:

[0079]

[0080] The second period:

[0081]

[0082]

[0083] The third period:

[0084]

[0085] Sum up the above three segment values:

[0086] 0.006194 + 0.01485 + 0.001516 ≈ 0.02256

[0087] Then divide by m = 3:

[0088]

[0089] This result indicates that for the comprehensive and collaborative perturbation intensity value of the current parameter in these three selected periods, it is approximately 0.00752. When this value is less than a certain stability threshold, it can be determined that the overall current fluctuation is small. If this value rises to a relatively large value, it indicates that the fluctuation intensity gradually increases and key monitoring should be carried out.

[0090] Set the stable value ranges for the current monitoring parameter, voltage monitoring parameter, and temperature monitoring parameter respectively according to the collaborative perturbation intensity value, determine whether the collaborative perturbation intensity value continuously exceeds the stable value range, and generate the abnormal parameter identification result. To execute this process, first compare the collaborative perturbation intensity values of the current, voltage, and temperature obtained from the previous step calculation with the stable intervals statistically obtained from the historical operation data. The stable value ranges can be set based on the long-term monitoring values of each parameter under normal load and full load conditions. For example, take the collaborative perturbation intensity interval of the current between 0.005 and 0.030, the collaborative perturbation intensity interval of the voltage between 0.004 and 0.025, and the collaborative perturbation intensity interval of the temperature between 0.006 and 0.028. These ranges are obtained through three months of continuous observation and data analysis. By comparing whether the specific {D1, D2, D3} is greater than or less than the upper or lower limits of their respective corresponding intervals, record the change situation of the collaborative perturbation intensity value in consecutive several periods in sequence. Then, according to the identification order of each period, count whether the same parameter exceeds the stable value range in all selected periods, sum up the number of periods with cumulative exceedance and write it into the current abnormal entry list, and then read the recorded entries for unified summary and finally generate the abnormal parameter identification result based on the statistical result.

[0091] The steps to obtain the parameter trend prediction result are as follows:

[0092] Select the parameters showing abnormal trends from the abnormal parameter identification result, including abnormal current, abnormal voltage, and abnormal temperature data, to obtain the selected list of abnormal parameters;

[0093] Based on the selected list of abnormal parameters, summarize the historical change patterns and the context in which abnormalities occur to obtain an integrated historical performance dataset;

[0094] According to the integrated historical performance dataset, apply linear regression to predict the future change trend of abnormal parameters, judge the change range and rate, and generate the parameter trend prediction result.

[0095] Specifically, according to the abnormal parameter identification results obtained previously, read the data records marked as abnormal current, abnormal voltage, and abnormal temperature. First, extract all abnormal items related to current from the abnormal parameter identification results and check their distribution on the operation time axis. For example, summarize the time periods when the current continuously exceeds the range of 0A to 5A. Then, perform the same extraction and induction for the voltage part. Compare the voltage with the range of 0V to 24V. If there are multiple records above 24V, merge these records into the voltage abnormality list to be reviewed. For the temperature aspect, also summarize according to the previously obtained temperature comparison range of 0°C to 90°C. Prioritize marking the records that exceed 90°C and have been confirmed as real high-temperature states. If the same parameter repeatedly shows abnormalities within one inspection cycle, put it into the key monitoring table additionally. Next, conduct a horizontal comparison of the three types of abnormal data of current, voltage, and temperature. Determine whether multiple types of abnormalities occur simultaneously at the same moment through the timestamp or record number. If multiple types of abnormalities occur concentratedly, add a combined label for convenient subsequent summarization. After completing this retrieval, a comprehensive list containing abnormal current, abnormal voltage, and abnormal temperature will be obtained. Attach the originally identified abnormal types to each record in this list to maintain consistency. Subsequently, further classify or segment and identify them according to their respective occurrence frequencies and fluctuation magnitudes. Store the abnormal records with obvious fluctuation trends and sporadic abnormal records in different data tables respectively. Finally, obtain the selected list of abnormal parameters by integrating the above extraction work.

[0096] Based on the selected list of abnormal parameters obtained previously, cross - retrieve the abnormal records that have been grouped into the same data table and summarize the corresponding historical change patterns. At the same time, extract the context conditions when the abnormality occurred from the previous operation logs or status tracking information, including on - site temperature, operating load, ventilation conditions, or installation location, etc. These information can be directly obtained from the pre - recorded operating environment and frequency observations. If it is found that a certain abnormality appears concentratedly during high - load periods, mark the high - load source in the record. Based on these details, correlate the same or similar abnormal situations with each other and integrate them into a complete historical evolution sequence in chronological order. Assume that there are several historical change patterns that can be mapped to this abnormal event currently. Then, for each numerical point saved in these patterns, retrieve them one by one and compare them with the abnormal record being processed. If the numerical fluctuation ranges are similar, merge them into the same pattern branch and arrange them in chronological order. If the numerical differences are large, mark the large difference value during integration for the next offset test. After completing these comparison tasks, a set of integrated historical performance datasets can be generated. At the end, form the historical pattern label corresponding to each abnormal record and the environmental context information at the abnormal time point.

[0097] According to the integrated historical performance datasets obtained previously, select the records with abnormal current, abnormal voltage, and abnormal temperature characteristics as inputs. Using the distribution of current values floating in the range of 0A to 5A, the dispersion of voltage values in the range of 0V to 24V, and the change curve of temperature values in the range of 0°C to 90°C as references, divide these data into training sets and validation sets in a certain batch. And during the training process, construct a linear regression model based on the time - series correlation relationships of current, voltage, and temperature in the aforementioned data. First, calculate the correlation coefficients between current, voltage, and temperature in each batch of training data, and then substitute these correlation coefficients into the equation to solve for the coefficients. If it is found that there is still an obvious problem of high error after several batches of training, re - label some high - fluctuation samples in combination with the error distribution table and make them weighted during subsequent iterations. After multiple iterative convergences, determine the final linear regression parameters and test the validation set. Compare the predicted current values, voltage values, and temperature values with the actual records to evaluate the deviation magnitude. If the deviation is within the pre - set allowable range, retain the model and infer the change amplitude and change rate of each parameter in the subsequent time period. When the estimation of the future - period trend of abnormal parameters is completed, a parameter trend prediction result will be generated.

[0098] The steps to obtain the equipment health assessment result are as follows:

[0099] Based on the parameter trend prediction result, evaluate the health impact degree of each parameter and calculate the equipment health score. The calculation formula is:

[0100]

[0101] Among them, H represents the device health score, J represents the number of parameters, P j represents the predicted value of the jth parameter, T j represents the target threshold of the jth parameter, S j represents the sensitivity coefficient of the jth parameter;

[0102] According to the device health score, determine whether the device health score is lower than the safety threshold. If it is lower, determine the device status as a potential risk and generate a device health assessment result.

[0103] Specifically, the advantage of the formula is that by simultaneously considering the deviation between the predicted value and the target threshold of each parameter and combining the sensitivity coefficient of each parameter, the comprehensive deviation degree between multiple measurement indicators and operation safety can be reflected in a unified expression;

[0104] The steps to obtain the J parameter are as follows: count the number of all predicted parameters currently involved in current, voltage, temperature, etc. For example, if 3 main parameters are involved, take J = 3;

[0105] P j The steps to obtain the parameter are as follows: read the predicted value of the jth parameter at a certain future time period from the parameter trend prediction result. For example, if the predicted current at a specific moment is 4.6A, the predicted voltage is 22.5V, and the predicted temperature is 85.0°C, etc., extract them according to each scenario;

[0106] T j The steps to obtain the parameter are as follows: combine industry standards or device manuals, etc., to set target thresholds for each monitored parameter. For example, set the current 5A, voltage 24V, and temperature 90°C as reference thresholds according to the rated load of the device, and record them as T1 = 5, T2 = 24, T3 = 90;

[0107] S j The steps to obtain the parameter are as follows: analyze the sensitivity of each parameter deviation based on a large amount of historical data. If a slight change in current has a greater impact on the device, its sensitivity coefficient can be set to a relatively higher value. For example, do the same processing for voltage and temperature based on long-term monitoring records, and finally form a functional trade-off value, such as S1 = 1.0, S2 = 1.5, S3 = 2.0;

[0108] Calculation process:

[0109] For example, if a device has three main parameters, take J = 3, obtain the predicted current value P1 = 4.6A, predicted voltage value P2 = 22.5V, predicted temperature value P3 = 85.0°C from the previous step, and combine the known target thresholds T1 = 5, T2 = 24, T3 = 90, and sensitivity coefficients S1 = 1.0, S2 = 1.5, S3 = 2.0 to perform multi-level operations:

[0110] Calculate the absolute deviation and its ratio:

[0111]

[0112] Add up the above ratios:

[0113]

[0114] Substitute into the formula:

[0115] H = 1 - exp(-3.9)

[0116] Calculate the exponential part:

[0117] exp(-3.9) ≈ 0.02024

[0118] Obtain the final H value:

[0119] H = 1 - 0.02024 = 0.97976

[0120] This result indicates that the device health score is approximately 0.97976 at this time. The closer the value is to 1, the lower the comprehensive deviation degree. If the predicted value or sensitivity coefficient is changed subsequently and this operation is performed again, a new health score can be obtained and the device status can be compared and analyzed.

[0121] Based on the device health score obtained previously, the corresponding scoring operations for each target value have been completed. Subsequently, the finally calculated health score is compared with the safety threshold recorded in the device management system. For example, the safety threshold is defined as 0.8, which is obtained by referring to multiple tests and industry experience and accumulated through actual operation. If the evaluated health score is less than 0.8, the device is marked as having potential risks. If the health score is higher than 0.8, the regular inspection process is maintained. At this time, various real-time and predicted data involved in the evaluation stage also need to be summarized and archived uniformly. If it is found that multiple device records are continuously in a low score state during the process, manual secondary confirmation is carried out in the records. For example, the current stability or temperature range of the device during the corresponding period is checked again to exclude sensor errors or special operating conditions, and finally the confirmation result is notified to the management personnel. If it is confirmed as a device with a long-term low score after repeated confirmation, it should be included in the subsequent maintenance scheduling plan, and thus the device health assessment result can be formed at the end of the overall process.

[0122] The steps to obtain the key influencing factor identification result are as follows:

[0123] Based on the device health assessment result, calculate the factor score of each index. The calculation formula is:

[0124]

[0125] Among them, F represents the factor score, Z is the normalization constant, K is the number of key performance indicators, and a k is the weight of the k-th indicator, and X k is the measured value of the k-th indicator, is the historical average value of the indicator, and S k is the standard deviation;

[0126] According to the factor scores, the performance indicator with the highest score is selected as the key influencing factor of the device performance, and the identification result of the key influencing factor is obtained.

[0127] Specifically, the advantage of the formula is that by measuring the deviation at the cubic level for each key performance indicator and integrating the weights and the normalization constant, the influence degree of each indicator on the device performance can be comprehensively considered under the same numerical structure;

[0128] The steps to obtain the Z parameter are as follows: Set the normalization reference value with reference to the deviation range of all key performance indicators. For example, count the upper and lower limits of the result distribution calculated by this formula from the data of multiple device tests, and then select an appropriate interval length value as the normalization constant;

[0129] The steps to obtain the K parameter are as follows: Count the total number of key performance indicators that need to be included in the factor analysis in the current device. For example, when there are 3 main indicators such as current fluctuation, voltage fluctuation, and temperature fluctuation, then let K = 3;

[0130] a k The steps to obtain the parameter are as follows: Evaluate and quantify the importance of each key performance indicator by combining historical failure data and laboratory test records. For example, count the influence frequencies of current, voltage, and temperature on device failures respectively in the test interval of the device running for 300 hours, and then assign different weights through distribution analysis;

[0131] X k The steps to obtain the parameter are as follows: Read the measured value of the k-th indicator at the current detection moment or the latest evaluation time period. For example, combine multiple sensor readings and manual secondary confirmation;

[0132] The steps to obtain the parameter are as follows: Collect the average values of each indicator in the historical data. For example, summarize the operating conditions in the past 6 months and calculate the average value;

[0133] S k The steps to obtain the parameter are as follows: Perform variance operation on the fluctuation interval of each indicator in the historical data and take the square root to obtain the standard deviation;

[0134] Calculation process:

[0135] For example, when K = 3 is set, and current, voltage, and temperature are selected as key performance indicators, and the weights are set as a1 = 0.35, a2 = 0.30, and a3 = 0.35, the measured current value X1 = 4.8 A, the measured voltage value X2 = 23.5 V, and the measured temperature value X3 = 85.2 °C, and the corresponding historical averages are respectively recorded as The corresponding standard deviations are S1 = 0.4, S2 = 1.2, and S3 = 2.0, and the normalization constant Z = 2.5;

[0136] Calculate the cube of the deviation of each indicator:

[0137] Current term:

[0138] Voltage term:

[0139] Temperature term:

[0140] Multiply by their respective weights and sum:

[0141]

[0142] Take the cube root of the sum result: (0.24494625) 1 / 3 ≈0.62;

[0143] Substitute the normalization constant Z = 2.5 and get F:

[0144] This result shows that the factor score is approximately 0.248 at this time. This value is used for horizontal comparison of the importance of multiple indicators in the device. The larger the value, the more likely the corresponding indicator is to be regarded as a key influencing factor;

[0145] According to the factor score results obtained above, they are associated with the equipment health assessment results. First, the measured data of each key performance indicator in the current monitoring cycle are summarized and compared with the historical average value one by one. These measured data include the floating frequency of the current value falling in the range of 0A to 5A, the voltage value comparison of the range of 0V to 24V, and whether the temperature value frequently exceeds the range of 0℃ to 90℃. By comparing, it can be identified which indicators have experienced more abnormal fluctuations or larger dispersion in the past period of time. Then, the factor scores of each indicator are sorted item by item. If the factor scores of some indicators exceed the distinction threshold obtained in advance from the multi-scenario test, they are judged as possible key influencing factors. When it is found that the scores of multiple indicators are relatively close, the measured original data of each indicator will be checked separately to further determine its fluctuation range. In this process, if an indicator appears in an extreme fluctuation state many times, a special mark will be added to the record so that it can be checked again in the subsequent analysis stage. Finally, the one with the highest score is classified as a key influencing factor and associated with the corresponding equipment components or operating environment elements. After the end, the key influencing factor identification result can be obtained.

[0146] The steps to obtain the maintenance strategy optimization results are:

[0147] Based on the identification results of key influencing factors, the corresponding equipment components, operating environment and operating parameters are classified and sorted to obtain a list of key maintenance elements;

[0148] According to the key maintenance element list, analyze the maintenance priority of each key maintenance element in each operating state, match the maintenance strategy, and form a maintenance task allocation plan;

[0149] Based on the maintenance task allocation plan, operational measures are formulated, corresponding maintenance operations are designed for each key maintenance element, and the maintenance measures are matched with the equipment operation plan to generate maintenance strategy optimization results.

[0150] Specifically, based on the key influencing factor identification results obtained previously, the associated equipment components, operating environments, and operating parameters are listed separately in a draft list. First, compare each of the previously identified high-impact indicators item by item, confirm the specific location of each influencing factor on the equipment or the corresponding operation link, and record the identification. For example, if it is found in actual monitoring that a certain bearing has too many times of temperature fluctuations exceeding the range of 0°C to 90°C, it is listed as a key concern area. If the corresponding environmental factors involve high site temperature or poor ventilation, they should also be written into the associated items. If there are records of current exceeding 5A or voltage exceeding 24V multiple times during a specific period in terms of operation, it indicates an overload risk and is included in the element list. Subsequently, classify item by item in combination with the previously recorded equipment model and rated specifications. The components prone to current anomalies are classified into the electrical maintenance group, the components and environments with high temperatures are put into the thermal management group, and the situations of improper operation are summarized into links such as training or process execution. If it is found during the classification process that some influencing factors can belong to multiple aspects simultaneously, additional cross-labels are added, and then check the relevant statistics filed previously to confirm whether this cross-label appears repeatedly. When all information is checked, a key maintenance element list is generated.

[0151] According to the key maintenance element list obtained previously, mark the maintenance priority of each key maintenance element under various operating states respectively. First, list several common states such as normal load conditions, high load conditions, and shutdown and maintenance conditions, and compare each record of the failure risk level or parameter fluctuation corresponding to the maintenance element item by item. For example, if the current remains within the range of 0A to 5A under normal load, it is temporarily recorded as medium priority. If the current exceeds 5A multiple times and the temperature is higher than 90°C under high load, it is marked as high priority and the source of the corresponding ambient temperature is recorded. If an operation error or component damage is still found for a certain element during shutdown and maintenance, it is regarded as a more urgent matter to be handled. Subsequently, compare the priority order of each element one by one and specify the corresponding maintenance method for each element. For example, in the electrical maintenance group, check the circuit connection or replace vulnerable components; in the thermal management group, strengthen the air cooling measures or check the blockage location of the heat dissipation ducts; in the operation execution link, provide special training for improper operations. After summarizing these priorities and maintenance methods item by item, a maintenance task assignment plan is formed.

[0152] Based on the maintenance task allocation plan obtained previously, specific operation measures are formulated item by item for key maintenance elements. For example, when the voltage is detected to be higher than 24V multiple times under the electrical maintenance group and the on-site temperature is close to 90°C, maintenance personnel need to be arranged in advance to synchronously check the line load and heat dissipation components. If there are multiple misoperations under high load in terms of operation execution, a duty officer can be arranged during this period to observe the process flow or set a certain alarm reference value in the control panel. When the monitored current exceeds 5A, a reminder will pop up on the operation interface to prompt the operator to immediately check the operating load status. For thermal management, focus on checking the circulating coolant or the smoothness of the air duct and adjusting the settings of the corresponding cooling equipment. Since the operation plan of the equipment has been scheduled in advance, it is necessary to check whether each maintenance operation will conflict with other operation links or production scheduling before implementation. If there are overlapping periods, priority will be given to arranging the maintenance of high-priority elements. Finally, after the corresponding measures for all key maintenance elements are compiled completely, they are integrated into a unified maintenance project schedule and synchronized to the equipment operation plan. When there are no contradictions after confirmation, the optimized result of the maintenance strategy can be output.

[0153] The steps to obtain the maintenance effect evaluation result are as follows:

[0154] According to the optimized result of the maintenance strategy, adjust, repair, or replace the components of the equipment, record the operation data and the changes in the equipment status during each maintenance process, and generate a maintenance operation record;

[0155] Based on the maintenance operation record, by monitoring the operation data of the equipment, including the change trends of current, voltage, and temperature, evaluate the operation performance of the equipment after maintenance, and generate a post-maintenance performance evaluation report;

[0156] Based on the post-maintenance performance evaluation report, calculate the improvement ratio of the equipment performance compared with that before maintenance, analyze the performance change trends of the equipment under different operating states, and obtain the maintenance effect evaluation result.

[0157] Specifically, according to the maintenance strategy optimization results obtained previously, when adjusting, repairing or replacing the components of the equipment, it is necessary to first retrieve the maintenance task allocation plan determined in the previous step and match the corresponding component information one by one. If the listed components have multiple records of current exceeding 5A or voltage greater than 24V in terms of electrical aspects, the equipment nameplate or rated parameter range should be checked again before formal operation. The current and voltage thresholds are based on previous operating data and industry standards. For example, the stable range during operation is defined between 0A and 5A and between 0V and 24V. If the value detected on-site by the maintenance personnel continues to exceed this range, a high-risk mark is added, and the corresponding list indicates which part has been replaced or repaired. For those thermal management components, the reference temperature is compared with the range of 0℃ to 90℃ and checked. Check whether the temperature repeatedly approaches or exceeds 90℃. If overheating occurs, disassemble the part to check the heat dissipation or add additional cooling materials. When the condition of the component is relatively minor, repair methods such as tightening the joints or replacing consumable parts can be used. If the component is severely worn after inspection, it can be replaced directly. During the operation, there is no need to record all the steps in multiple locations. Instead, the disassembly and assembly time, inspection results and installation status are recorded in the order of execution. Finally, the entire maintenance operation process is summarized in the same database at one time with key time nodes and whether the equipment status returns to the allowable range after each specific operation. For example, reconfirm that the current is stable at 0A to 5A, the voltage is maintained at 0V to 24V, or the temperature does not fluctuate abnormally. When all adjustments or repair items are completed, a maintenance operation record can be generated.

[0158] According to the maintenance operation records obtained above, the real-time data of current, voltage and temperature are read and compared item by item. If the current fluctuation is maintained in the range of 0A to 5A during a period of operation, it means that the current load is relatively stable. If the voltage stays in the range of 0V to 24V for most of the time, it is considered that there is no obvious deviation. For temperature, the data collected on site are compared with the range of 0℃ to 90℃ one by one. If the monitoring results show that the temperature is still in a relatively normal range after several hours of continuous operation and does not exceed 90℃ many times, it will be marked in the record as maintaining normal. If it is found during the detection at a certain time If the segment temperature is only a small difference from 90°C, it is necessary to make a note in combination with the operation frequency or the ambient temperature and verify again whether there is insufficient ventilation or excessive load. After the comparison of the current, voltage, and temperature values ​​is completed, the mean and peak values ​​of each parameter in the same time slice are statistically analyzed, and the range of change is obtained by horizontally comparing these values ​​with the data of the previous maintenance operation stage, and the deviation intervals are recorded one by one in the same comprehensive table. If the deviation is higher than the historical reference standard, it is necessary to judge the possible omission risks that may still exist after the component is repaired. Finally, after these value comparison and registration steps are completed, a post-maintenance performance evaluation report is generated.

[0159] According to the post-maintenance performance evaluation report obtained above, the current, voltage and temperature operating data of the equipment at different time periods before and after maintenance are summarized and compared. Then, the performance under typical operating conditions such as normal load, high load and low load are selected and sorted out. Whether the current exceeds 5A multiple times under high load conditions before maintenance and whether it drops to less than 5A after maintenance is counted item by item. For temperature, check whether it exceeds 90℃ under high load in the same way. If it basically remains below 90℃ after maintenance, record the corresponding state, and check whether the voltage is continuously stable within 24V in the same time period. If there are multiple records of more than 24V before maintenance, list them together for comparison. Then, perform differential or percentage calculation on these before and after data to obtain the improvement ratio. For example, in the normal load scenario, the average current value was originally 4.2A, and the record after maintenance was 3.8A. The difference is divided by the original average value and converted into a percentage. When all operating conditions are compared, the performance change trend under each scenario can be obtained. If the improvement ratio of some scenarios is particularly obvious, it is marked separately in the summary part. Finally, the maintenance effect evaluation result is obtained based on the results of these calculations and comparisons.

[0160] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A diagnostic method for the health status of electromechanical equipment based on machine learning, characterized in that, Including the following steps: Collect the current, voltage and temperature parameters of the device. Through numerical processing, calculate the real-time average value and standard deviation of each parameter to obtain the parameter monitoring result; Based on the parameter monitoring result, determine whether the parameter is abnormal and generate an abnormal parameter identification result; Based on the abnormal parameter identification result, select parameters for analysis. Combine past performance data and use a regression algorithm to predict the change trend of the parameter in the future time period and generate a parameter trend prediction result; Evaluate the parameter trend prediction result, judge the health status of the device, and obtain the device health assessment result; Based on the device health assessment result, identify the key influencing factors of the device performance, extract the key influencing factors, and obtain the key influencing factor identification result; Adjust the maintenance strategy according to the key influencing factor identification result, formulate targeted maintenance measures, and generate an optimized maintenance strategy result; Based on the optimized maintenance strategy result, implement the maintenance measures, monitor the maintenance effect through real-time data, and calculate the performance improvement ratio of the device after maintenance to obtain the maintenance effect assessment result.

2. The diagnostic method for the health state of electromechanical equipment based on machine learning according to claim 1, characterized in that The steps for obtaining the parameter monitoring result are as follows: Collect the current, voltage and temperature data of the device through sensors, record the current value, voltage value and temperature value of each collection to obtain the original collection data; According to the original collection data, calculate the real-time average value and standard deviation of the current data, the real-time average value and standard deviation of the voltage data, and the real-time average value and standard deviation of the temperature data to generate a real-time monitoring result; Based on the real-time monitoring result, integrate the real-time monitoring results of the current data, the real-time monitoring results of the voltage data and the real-time monitoring results of the temperature data to form a parameter monitoring result.

3. The diagnostic method for the health status of electromechanical equipment based on machine learning according to claim 1, wherein The steps for obtaining the abnormal parameter identification result are as follows: Based on the parameter monitoring result, intercept the current monitoring numerical sequence, voltage monitoring numerical sequence and temperature monitoring numerical sequence within the most recent 30 days, divide them into fixed equal-width time periods by day respectively, and extract the maximum value, minimum value, first and last difference and the number of monotonic intervals of each time period to generate a current fluctuation profile group, a voltage fluctuation profile group and a temperature fluctuation profile group; According to the current fluctuation profile group, voltage fluctuation profile group and temperature fluctuation profile group, calculate the co-disturbance intensity value. The calculation formula is: Among them, D c is the co-perturbation intensity value of the c-th type of parameter, A cu is the maximum value in the u-th period, B cu is the minimum value in the u-th period, C cu is the head and tail difference, Z cu is the number of consecutive segments that are monotonically increasing or decreasing within the u-th period, Y cu is the direction switching frequency between two adjacent peaks, and m is the total number of periods for statistics; Set the stable value range for the current monitoring parameter, voltage monitoring parameter and temperature monitoring parameter respectively according to the co-disturbance intensity value, and judge whether the co-disturbance intensity value continuously exceeds the stable value range to generate an abnormal parameter identification result.

4. The diagnostic method for the health state of the electromechanical device based on machine learning according to claim 1, characterized in that The steps for obtaining the parameter trend prediction result are as follows: Select the parameters showing abnormal trends from the abnormal parameter identification result, including abnormal current, abnormal voltage and abnormal temperature data, to obtain the selected list of abnormal parameters; Based on the selected list of abnormal parameters, summarize the historical change patterns and the context of abnormal occurrences to obtain an integrated historical performance dataset; According to the integrated historical performance dataset, apply linear regression to predict the future change trend of the abnormal parameter, judge the change range and rate, and generate a parameter trend prediction result.

5. The diagnostic method for the health status of electromechanical equipment based on machine learning according to claim 1, wherein The steps for obtaining the device health assessment result are as follows: Based on the parameter trend prediction results, evaluate the health impact degree of each parameter, and calculate the device health score. The calculation formula is as follows: Among them, H represents the device health score, J represents the number of parameters, P j represents the predicted value of the j-th parameter, T j represents the target threshold of the j-th parameter, S j represents the sensitivity coefficient of the j-th parameter; According to the device health score, determine whether the device health score is lower than the safety threshold. If it is lower, determine the device status as a potential risk and generate a device health assessment result.

6. The diagnostic method for the health status of electromechanical equipment based on machine learning according to claim 1, characterized in that, The steps for obtaining the key influencing factor identification results are as follows: Based on the device health assessment results, calculate the factor scores of each indicator; According to the factor scores, select the performance indicator with the highest score as the key influencing factor of the device performance to obtain the key influencing factor identification results.

7. The diagnostic method for the health status of electromechanical equipment based on machine learning according to claim 1, characterized in that, The steps for obtaining the maintenance strategy optimization results are as follows: Based on the key influencing factor identification results, classify and organize the corresponding device components, operating environments, and operating parameters to obtain a list of key maintenance elements; According to the list of key maintenance elements, analyze the maintenance priorities of each key maintenance element in each operating state and match the maintenance strategies to form a maintenance task allocation plan; Based on the maintenance task allocation plan, formulate operation measures, design corresponding maintenance operations for each key maintenance element, and match the maintenance measures with the device operation plan to generate maintenance strategy optimization results.

8. The diagnostic method for the health state of the electromechanical equipment based on machine learning according to claim 1, characterized in that, The steps for obtaining the maintenance effect evaluation results are as follows: According to the maintenance strategy optimization results, adjust, repair, or replace the components of the device, record the operation data and the changes in the device status during each maintenance process, and generate a maintenance operation record; Based on the maintenance operation record, evaluate the operation performance of the device after maintenance by monitoring the operation data of the device, including the change trends of current, voltage, and temperature, and generate a post-maintenance performance evaluation report; Based on the post-maintenance performance evaluation report, calculate the improvement ratio of the device performance compared with that before maintenance, analyze the performance change trends of the device in different operating states, and obtain the maintenance effect evaluation results.

9. The diagnostic device for the diagnostic method of the health state of the electromechanical device based on machine learning according to any one of claims 1-8, characterized in that, Including: A data acquisition module that collects the current, voltage, and temperature data of the motor, calculates the real-time average value and standard deviation of the data, and generates parameter monitoring results; An anomaly analysis module that, based on the parameter monitoring results, compares each item of data with a preset threshold, identifies the data that exceeds the normal range, and generates anomaly parameter identification results; A trend prediction module that uses the anomaly parameter identification results to analyze the historical data of related parameters, predicts the change trends within a future time period, and generates parameter trend prediction results; A health assessment module that analyzes the parameter trend prediction results, judges the current health status of the device, and generates a device health assessment result; A maintenance strategy optimization module that, based on the device health assessment results, analyzes the key influencing factors of the device performance, extracts and analyzes the key influencing factors, formulates or adjusts the maintenance measures of the device, and generates maintenance strategy optimization results.

Images