Intelligent inspection system for power equipment based on artificial intelligence

Abstract

The invention relates to the technical field of industrial equipment management, in particular to an intelligent power equipment inspection system based on artificial intelligence, which comprises a fault mode generation module, a fault detection module, a fault classification module, an inspection feature optimization module and an anomaly prediction and evaluation module. According to the method, the comprehensiveness of equipment state sensing is optimized by collecting the multi-dimensional operation parameters of the power equipment and analyzing the signal amplitude distribution and frequency domain change, the potential fault features are accurately recognized by combining the dynamic feature range and the signal distribution deviation matching, the signal anomaly detection accuracy is improved, and the fault detection accuracy is improved. The method comprises the following steps: extracting differentiated characteristic values of current and temperature signals, optimizing fault type classification, improving the positioning capability of a problem root, dynamically extracting and matching a signal trend, generating an inspection parameter matrix, realizing global dynamic monitoring of an operation state, and predicting a model based on a signal change rate and a time node range. The real-time performance and precision of abnormal trend prediction are improved, and the reliability of equipment operation and maintenance is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment management, and in particular to an intelligent inspection system for power equipment based on artificial intelligence. Background Art

[0002] The field of industrial equipment management technology includes technologies for real-time monitoring and analysis of the operating status of industrial production equipment, fault diagnosis, status prediction, and operation and maintenance optimization. The core content of this technology is to collect equipment operating parameters through sensors and combine data processing technology to monitor and evaluate the operating status of equipment. Systematically speaking, industrial equipment management covers a series of tasks from the collection and storage of equipment operating data, status evaluation and fault diagnosis, to maintenance and optimization decisions. Its purpose is to ensure the safe and stable operation of equipment, improve production efficiency and extend the service life of equipment.

[0003] Among them, the AI-based intelligent inspection system for power equipment refers to a system that uses AI technology to automatically inspect and monitor the status of power equipment. The system targets the real-time status perception and potential fault analysis of power equipment, including extracting the appearance features of power equipment based on image recognition technology, detecting abnormalities in the operating parameters of power equipment based on data analysis technology, and conducting comprehensive analysis of the operating status of equipment through knowledge graph technology. The system mainly relies on the integration and analysis of multi-dimensional power equipment data to achieve intelligent inspection and comprehensive evaluation of equipment status.

[0004] Existing technologies rely on single parameter monitoring for equipment status perception and are unable to conduct comprehensive multi-dimensional data analysis, resulting in insufficient identification coverage of potential problems. Static data models cannot capture the dynamic changes of operating signals, and diagnosis is slow to respond to real-time status. Fixed parameter ranges limit the effective identification of non-standardized anomalies and are prone to missing potential faults in complex scenarios. Image recognition and single analysis methods are easily affected by external environmental interference, and data collection stability and accuracy are insufficient, which affects the effectiveness of equipment status assessment, reduces the reliability of anomaly prediction, and makes it difficult to support the efficient operation and maintenance needs of complex power equipment. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent inspection system for power equipment based on artificial intelligence.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: The intelligent inspection system of power equipment based on artificial intelligence includes: The fault mode generation module obtains the operating parameters of the power equipment through the inspection robot, inputs the current value, temperature value and vibration signal value, counts the amplitude distribution of the signal, extracts the frequency domain characteristics and change rate of the signal, calculates the dynamic characteristic range of the signal, establishes the fault characteristic mapping model, adjusts the parameters to match the signal distribution deviation, and generates the fault mode distribution value; The fault detection module obtains the current signal offset value and the frequency domain deviation in the inspection signal based on the fault mode distribution value, compares the interval difference between the inspection signal and the normal operating parameters, calculates the amplitude difference rate of the signal abnormality, establishes a fault feature deviation model, and generates the abnormal signal amplitude ratio; The fault classification module analyzes the characteristic values ​​of the current signal and the temperature signal in the differentiation interval based on the abnormal signal amplitude ratio, compares the overlap between the signal characteristic values ​​and the fault category interval, establishes a classification feature matrix, screens the signal contribution rate, and generates a fault type distribution coefficient; The inspection feature optimization module extracts the local trend value and global change rate of the current and temperature signals based on the fault type distribution coefficient, normalizes the signal category feature weight, dynamically matches the signal trend, extracts the inspection parameter matrix, and calculates the inspection feature weight distribution value; The abnormal prediction evaluation module analyzes the change rates of vibration signals and temperature signals based on the inspection feature weight distribution values, establishes a dynamic trend matrix, calculates the signal change range of time nodes, establishes a time distribution prediction model, and generates abnormal trend prediction values.

[0007] The fault mode distribution value is specifically the signal amplitude distribution, signal frequency domain characteristics and change rate, and signal dynamic feature range. The fault feature deviation model includes the current signal offset value, frequency domain deviation, and signal abnormal amplitude difference rate. The abnormal signal amplitude ratio includes the current signal characteristic value, the temperature signal characteristic value, and the overlap between the signal characteristic value and the fault category interval. The fault type distribution coefficient is specifically the classification feature matrix and the signal contribution rate. The inspection feature weight distribution value is specifically the local trend value, the global change rate, and the signal category feature weight. The abnormal trend prediction value includes the vibration signal change rate, the temperature signal change rate, and the time node signal change range.

[0008] As a further solution of the present invention, the step of obtaining the failure mode distribution value is specifically as follows: Extract the amplitude distribution of current value, temperature value and vibration signal value, calculate the change rate of multiple signal amplitudes in the frequency domain, and generate amplitude distribution change rate parameters; Call the amplitude distribution change rate parameter, compare its value to see if it exceeds the dynamic feature range according to the upper and lower limits, classify the signal amplitude according to the frequency domain characteristics, extract the signal feature change value that meets the conditions of the upper and lower limits of the dynamic feature range, and calculate and generate the signal change rate within the dynamic feature range; The signal change rate within the dynamic feature range is called, the distribution deviation is analyzed, and the signal characteristic value mapping relationship is corrected based on the adjustment parameters, using the formula: ; Calculate and generate failure mode distribution values; in, represents the failure mode distribution value, Represents the rate of change of the signal characteristic value, represents the weight of the signal amplitude, Represents the lower limit of the dynamic feature range, Represents the upper limit of the dynamic feature range, represents the weight coefficient of the adjustment parameter, represents the signal feature mapping value, represents the mean value of signal characteristics, Represents the number of signal feature points, Represents the number of signal mapping points, , They are the indexes of signal feature points and mapping points respectively.

[0009] As a further solution of the present invention, the step of obtaining the abnormal signal amplitude ratio is specifically as follows: According to the fault mode distribution value, the offset value and frequency domain deviation of the current signal in the inspection signal are obtained, the frequency domain deviation is used to statistically calculate the variation range of the differentiated signal, the offset value and the variation range are classified and sorted, and the current signal deviation data is generated; Calling the current signal deviation data, comparing the interval difference between the inspection signal and the normal operating parameters, calculating the deviation amplitude of the inspection signal in multiple intervals, analyzing the variation trend of the deviation amplitude, and counting the amplitude difference rate of the signal abnormality to generate signal abnormality amplitude difference data; Using the signal abnormal amplitude difference data, a fault feature deviation model is established using the formula: ; Generate abnormal signal amplitude ratio; in, represents the overall proportion of signal deviation, Representative The amplitude of the signal observation, Representative The normal amplitude reference of a signal, represents the number of all signal data points, is the index of the signal data point.

[0010] As a further solution of the present invention, the step of obtaining the fault type distribution coefficient is specifically as follows: Based on the abnormal signal amplitude ratio, the characteristic values ​​of the current signal and the temperature signal in the differentiated interval are analyzed, and the fluctuation values ​​and frequency domain characteristics of the signals in the differentiated interval are classified item by item to generate differentiated characteristic data; The differentiated feature data is called, the overlap between the signal feature value and the fault category interval is compared, and the criticality of the multi-category interval is analyzed by counting and sorting the contribution rate of the signal amplitude in the overlapping interval and combining the signal feature weight parameter to generate the signal contribution rate data; Using the signal contribution rate data, a classification feature matrix is ​​established, and the weight relationship between the feature distribution of the differentiated signal category and the fault category interval is calculated using the formula: ; Calculate and generate the fault type distribution coefficient; in, represents the fault type distribution coefficient, Representative Class signal contribution rate, Representative The weight of the class signal, Representative The normal baseline contribution rate of the class signal, Representative Class The amplitude contribution of each interval is Represents the number of signal categories, represents the number of intervals, is the signal category index, is the interval index.

[0011] As a further solution of the present invention, the step of obtaining the inspection feature weight distribution value is specifically as follows: Based on the fault type distribution coefficient, extract the local trend value and global change rate of the current and temperature signals, and generate local and global signal analysis data by analyzing the fluctuation characteristics of the signals in the local and global ranges item by item and combining the change law of the signal trend; Calling the local and global signal analysis data, normalizing the signal category feature weights, normalizing the signal categories to a unified range by calculating the ratio of multi-category signal feature values, and allocating feature weights of multi-category signals to generate normalized signal weight data; Using the normalized signal weight data, dynamically matching the signal trend, and extracting the inspection parameter matrix, the formula is: ; Calculate and generate inspection feature weight distribution values; in, Represents the inspection feature weight distribution value, Representative The local trend value of the class signal, Representative The global rate of change of the class signal, Representative The normalized weight of the class signal, Representative Stability index of class signal, Represents the number of signal categories, Signal category index.

[0012] As a further solution of the present invention, the step of obtaining the abnormal trend prediction value is specifically as follows: Based on the inspection feature weight distribution value, the change rate of the vibration signal and the temperature signal is analyzed, and the dynamic change rate of each signal is extracted by calculating the change ratio between the signal fluctuation amplitude and the time interval, and combining the signal feature parameters to generate the change rate data; Call the change rate data, establish a dynamic trend matrix, calculate the change amplitude and distribution law of the signal at the differentiated time nodes, and extract the signal change range data at the time nodes in combination with the signal fluctuation characteristics; Using the signal change range data of the time nodes, a time distribution prediction model is established, combining the dynamic change of the signal with the time node distribution, and using the formula: ; Calculate and generate abnormal trend prediction values; in, represents the abnormal trend prediction value, Representative The rate of change of the signal, Representative The time node value of the class signal, represents the mean value of the time node, Represents the standard deviation of the time node, Representative The strength value of the signal, Represents the number of signal categories, Represents the number of time nodes, , They are the indexes of signal category and time node respectively.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by collecting multi-dimensional operating parameters of power equipment, including current values, temperature values, and vibration signal values, and analyzing signal amplitude distribution and frequency domain changes, the comprehensiveness of equipment status perception is optimized. Combined with the dynamic feature range and signal distribution deviation matching, potential fault characteristics are accurately identified and the accuracy of signal anomaly detection is improved. By extracting differentiated characteristic values ​​of current and temperature signals, the fault type classification is optimized and the ability to locate the root cause of the problem is improved. Dynamically extract and match signal trends, generate an inspection parameter matrix, and realize global dynamic monitoring of the operating status. Based on the prediction model of signal change rate and time node range, the real-time and accuracy of abnormal trend prediction is improved, and the reliability of equipment operation and maintenance is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a system flow chart of the present invention; Figure 2 A flowchart of the steps for obtaining the fault mode distribution value of the present invention; Figure 3 Flow chart of the steps of obtaining the abnormal signal amplitude ratio of the present invention; Figure 4 A flow chart of the steps for obtaining the fault type distribution coefficient of the present invention; Figure 5 This is a flow chart of the steps for obtaining the inspection feature weight distribution value of the present invention; Figure 6 The present invention is a flowchart of the steps for obtaining the abnormal trend prediction value. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with 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 intended to limit the present invention.

[0016] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0017] Example 1: Please refer to Figure 1 , the intelligent inspection system for power equipment based on artificial intelligence includes: The fault mode generation module obtains the operating parameters of the power equipment through the inspection robot, inputs the current value, temperature value and vibration signal value, counts the amplitude distribution of the signal, extracts the frequency domain characteristics and change rate of the signal, calculates the dynamic characteristic range of the signal, establishes the fault characteristic mapping model, adjusts the parameters to match the signal distribution deviation, and generates the fault mode distribution value; The fault detection module obtains the current signal offset value and frequency domain deviation in the inspection signal based on the fault mode distribution value, compares the interval difference between the inspection signal and the normal operating parameters, calculates the amplitude difference rate of the signal abnormality, establishes the fault feature deviation model, and generates the abnormal signal amplitude ratio; The fault classification module analyzes the characteristic values ​​of the current signal and the temperature signal in the differentiated interval based on the abnormal signal amplitude ratio, compares the overlap between the signal characteristic values ​​and the fault category interval, establishes a classification feature matrix, screens the signal contribution rate, and generates the fault type distribution coefficient; The inspection feature optimization module extracts the local trend value and global change rate of the current and temperature signals based on the fault type distribution coefficient, normalizes the signal category feature weight, dynamically matches the signal trend, extracts the inspection parameter matrix, and calculates the inspection feature weight distribution value; The abnormal prediction evaluation module analyzes the change rate of vibration signals and temperature signals based on the inspection feature weight distribution value, establishes a dynamic trend matrix, calculates the signal change range of time nodes, establishes a time distribution prediction model, and generates abnormal trend prediction values.

[0018] The fault mode distribution values ​​are specifically the signal amplitude distribution, signal frequency domain characteristics and change rate, and signal dynamic feature range. The fault feature deviation model includes the current signal offset value, frequency domain deviation, and signal abnormal amplitude difference rate. The abnormal signal amplitude ratio includes the current signal characteristic value, temperature signal characteristic value, and the overlap between the signal characteristic value and the fault category interval. The fault type distribution coefficient is specifically the classification feature matrix and the signal contribution rate. The inspection feature weight distribution values ​​are specifically the local trend value, the global change rate, and the signal category feature weight. The abnormal trend prediction value includes the vibration signal change rate, the temperature signal change rate, and the time node signal change range.

[0019] See also Figure 2 , the specific steps for obtaining the failure mode distribution value are: Extract the amplitude distribution of current value, temperature value and vibration signal value, calculate the change rate of multiple signal amplitudes in the frequency domain, and generate amplitude distribution change rate parameters; The numerical range of each signal type is screened, and its segmented distribution is classified into multiple intervals according to the amplitude. The average amplitude and standard deviation of each segment are calculated one by one. The calculation results are used to draw an amplitude distribution diagram to show the frequency domain characteristic distribution and change rate of the signal and obtain the amplitude distribution change rate parameters.

[0020] Call the amplitude distribution change rate parameter, compare its value with the upper and lower limits to see if it exceeds the dynamic feature range, classify the signal amplitude according to the frequency domain characteristics, extract the signal feature change values ​​that meet the upper and lower limits of the dynamic feature range, and calculate the signal change rate within the dynamic feature range; By setting the upper and lower limits of the dynamic feature range as reference values, the signal amplitude distribution is gradually screened to see if it exceeds the upper and lower limits. The screened signal feature change value is compared with the dynamic feature range to extract the signal change rate within the dynamic range. Each signal value that meets the range is accumulated item by item to calculate the mean change rate, and the signal change rate within the dynamic feature range is generated by combining the cumulative amplitude change rate.

[0021] Call the signal change rate within the dynamic feature range, analyze the distribution deviation, and correct the signal feature value mapping relationship based on the adjustment parameters, using the formula: ; Calculate and generate failure mode distribution values; Among them, represents the failure mode distribution value, Represents the rate of change of the signal characteristic value, represents the weight of the signal amplitude, Represents the lower limit of the dynamic feature range, Represents the upper limit of the dynamic feature range, represents the weight coefficient of the adjustment parameter, represents the signal feature mapping value, represents the mean value of signal characteristics, Represents the number of signal feature points, Represents the number of signal mapping points, , They are the indexes of signal feature points and mapping points respectively.

[0022] formula: ; The benefit of the formula is that by simultaneously introducing the signal amplitude change rate, weight, dynamic feature range and signal feature mapping value into the calculation, the dynamic change characteristics of the signal can be comprehensively evaluated, and the accuracy of the signal distribution model can be significantly improved.

[0023] Detailed explanation of the formula and the process of formula calculation and derivation: ; Calculate the denominator: ; Calculate the numerator: ; Calculate the fractional term: ; Compute the summation term: ; Compute the squared term: ; Compute the sum of squares: ; Compute the square root term: ; Calculate the weighted terms: ; Calculate the final result: ; The results show that the fault mode distribution value of the signal is 2.191, which is directly related to the amplitude distribution change rate of the signal, the change rate within the dynamic characteristic range and the mapping characteristic value. This value can effectively characterize the signal distribution characteristics and can be further used to generate a fault feature mapping model.

[0024] See also Figure 3 , the specific steps for obtaining the abnormal signal amplitude ratio are: According to the fault mode distribution value, the offset value and frequency domain deviation of the current signal in the inspection signal are obtained, the frequency domain deviation is used to statistically calculate the variation range of the differentiated signal, the offset value and variation range are classified and sorted, and the current signal deviation data is generated; The signal offset value and frequency domain deviation in the fault mode distribution data are used to perform segmented analysis on the signal offset value, and the frequency domain deviation is compared with the preset reference signal to determine the offset amplitude range. The relative changes of the offset value and the reference signal in each frequency domain segment are calculated one by one, and the change amplitude in each frequency domain segment is quantified. The frequency domain segment average calculation formula is used: ,in, represents the average offset in the frequency domain, Represents the offset value in the signal frequency domain segment, represents the reference frequency domain segment value of the signal, Represents the number of frequency domain segments. This calculation formula is used to process all offset signals segment by segment, and the current signal deviation data is generated through the normalization method.

[0025] Call the current signal deviation data, compare the interval difference between the inspection signal and the normal operating parameters, calculate the deviation amplitude of the inspection signal in multiple intervals, analyze the change trend of the deviation amplitude, count the amplitude difference rate of signal abnormality, and generate signal abnormality amplitude difference data; The interval differences between the signal deviation value and the normal operating parameters are analyzed by segment classification. The interval data with larger deviation values ​​are extracted according to the classification results. The change rate of the deviation amplitude within the deviation interval is calculated, and the change rate calculation formula is adopted: ,in, represents the rate of change of deviation, Represents the absolute change in the deviation amplitude, Represents the average amplitude of the reference signal. The change rate is calculated for each interval and combined to form a global signal deviation change rate curve to generate signal abnormal amplitude difference data.

[0026] Using the signal abnormal amplitude difference data, a fault feature deviation model is established using the formula: ; Generate abnormal signal amplitude ratio; in, represents the overall proportion of signal deviation, Representative The amplitude of the signal observation, Representative The normal amplitude reference of a signal, represents the number of all signal data points, is the index of the signal data point.

[0027] formula: ; The benefit of the formula is that by introducing the square sum of relative differences in signal amplitude for normalization, the sensitivity of deviation calculation to abnormal signal distribution is enhanced, and the accuracy of signal deviation quantification and the robustness of calculation are improved; Detailed explanation of the formula and the process of formula calculation and derivation: in, represents the overall proportion of signal deviation, Representative The amplitude of the signal observation, Representative The normal amplitude reference of a signal, represents the number of all signal data points, is the index of the signal data point, assuming that the observed signal amplitude is , the reference signal amplitude is , put it into the formula: ; Calculate step by step: ; ; ; The result shows that the abnormal signal amplitude ratio is 0.0295, indicating that the degree of signal deviation is low. The cause and distribution characteristics of the signal anomaly can be further analyzed in combination with other data to generate the abnormal signal amplitude ratio.

[0028] See also Figure 4 , the specific steps for obtaining the fault type distribution coefficient are: Based on the abnormal signal amplitude ratio, the characteristic values ​​of the current signal and the temperature signal in the differentiated interval are analyzed, and the fluctuation value and frequency domain characteristics of the signal in the differentiated interval are classified item by item to generate differentiated characteristic data; The numerical value in the abnormal signal amplitude ratio is called as a benchmark, and it is cross-compared with the characteristic signal in the differentiated interval to identify the interval range with significant signal differences. The amplitude fluctuation parameters and time series characteristic values ​​of the current signal and temperature signal in each difference interval are extracted respectively, and normalization processing is performed based on the variation range of the difference amplitude to standardize the statistical benchmark values ​​of each signal. Subsequently, the normalization results are classified using the frequency domain distribution characteristics in the time series, and the difference signals are mapped to different characteristic value intervals. The signal amplitude fluctuations in each interval are classified and counted to generate differentiated characteristic data.

[0029] Call the differentiated feature data, compare the overlap between the signal feature value and the fault category interval, and generate signal contribution rate data by counting and sorting the signal amplitude contribution rate in the overlapping interval and combining the signal feature weight parameter to analyze the criticality of the multi-category interval; The signal characteristic value is judged by the characteristic boundary range of the fault category interval to determine whether each signal category is in the overlapping interval. The contribution rate of the signal amplitude in each overlapping interval is counted, and the amplitude contribution rate of the signal category is classified and summarized. The distribution range and central value of the amplitude of the category with a larger contribution rate are calculated, and the central value is adjusted in combination with the characteristic weight parameter of the signal category. The overall contribution rate of each signal category is obtained by statistics to generate signal contribution rate data.

[0030] Using the signal contribution rate data, a classification feature matrix is ​​established. By calculating the weight relationship between the feature distribution of the differentiated signal category and the fault category interval, the formula is used: ; Calculate and generate the fault type distribution coefficient; in, represents the fault type distribution coefficient, Representative Class signal contribution rate, Representative The weight of the class signal, Representative The normal baseline contribution rate of the class signal, Representative Class The amplitude contribution of each interval is Represents the number of signal categories, represents the number of intervals, is the signal category index, is the interval index.

[0031] formula: ; The benefit of the formula is that by weighting the signal class parameter Sum signal contribution Combined calculation, combined with the normal benchmark contribution rate The logarithmic deviation and absolute value calculations improve the ability to distinguish the distribution of fault category characteristics, thereby optimizing the accuracy of fault classification; Detailed explanation of the formula and the process of formula calculation and derivation: Set the number of signal categories , the number of intervals ; Signal contribution rate , , ; The weight corresponding to the signal category , , ; Corresponding normal benchmark contribution rate , , ; The total contribution rate of each interval , , ; Specific calculation process: 1. Calculate the logarithmic deviation: ; ; ; 2. Calculate the absolute value of the logarithm and multiply the weight: ; ; ; 3. Take the square root after summing: ; ; ; The results show that the fault type distribution coefficient It indicates that the contribution rate of signal categories within the interval is significantly different. After the weight and bias of the classification feature matrix are adjusted, the fault classification can more accurately distinguish the distribution of signal category features.

[0032] See also Figure 5 ,The specific steps for obtaining the inspection feature weight distribution value are: Based on the fault type distribution coefficient, the local trend value and global change rate of the current and temperature signals are extracted. By analyzing the fluctuation characteristics of the signals in the local and global ranges item by item and combining the change law of the signal trend, local and global signal analysis data are generated; By extracting the local trend value and global change rate of the current and temperature signals, according to the fluctuation characteristics of the signals in different time periods, the maximum value, minimum value and average change rate of each signal in a fixed time interval are first extracted, and then the fluctuation range of the signal is segmented and classified. The local trend value is calculated by segmented statistics, and the global change rate is calculated after comparing it with the global signal change range. The global change rate is obtained by dividing the total change of the signal by the number of samples in the complete time period. At the same time, the classified local trend value and the global change rate are cross-compared. The change pattern of the signal is analyzed by superposition of the two to generate local and global signal analysis data.

[0033] Call local and global signal analysis data, normalize signal category feature weights, calculate the proportion of multi-category signal feature values, normalize signal categories to a unified range, and assign feature weights of multi-category signals to generate normalized signal weight data; First, the feature weights of each signal category are proportionally divided, and the feature values ​​are normalized to the interval between 0 and 1 according to the total weight within the category, using the normalization formula: , calculate the normalized weight, where Represents the original weight of the signal, normalized weight After calculation, they will be redistributed by category, and the weights will be dynamically adjusted according to the ratio of signal fluctuation frequency and total number of samples, through the dynamic adjustment formula: ,middle The fluctuation frequency of the signal is combined with the characteristic weight of the normalized signal and the dynamic weight adjustment to generate the normalized signal weight data.

[0034] Use normalized signal weight data, dynamically match signal trends, and extract inspection parameter matrix using the formula: ; Calculate and generate inspection feature weight distribution values; in, Represents the inspection feature weight distribution value, Representative The local trend value of the class signal, Representative The global rate of change of the class signal, Representative The normalized weight of the class signal, Representative Stability index of class signal, Represents the number of signal categories, Signal category index.

[0035] formula: ; The benefit of the formula is that, through the combination of local trend value, global change rate, normalized weight and signal stability factor, it can dynamically reflect the contribution of different signal categories to the overall inspection parameter matrix and improve the accuracy of signal weight distribution; Detailed explanation of the formula and the process of formula calculation and derivation: Set the local trend value , global rate of change , normalized weight , signal stability index , substitute the above parameters into the formula: ; Calculate item by item: ; ; ; ; ; The result shows that the inspection feature weight distribution value is 0.339, which reflects the overall weight distribution characteristics of each signal category in the inspection parameter matrix, and is used for the comprehensive evaluation and trend analysis of subsequent inspection parameters.

[0036] See also Figure 6 , the specific steps for obtaining the abnormal trend prediction value are: Based on the inspection feature weight distribution value, the change rate of the vibration signal and the temperature signal is analyzed. By calculating the change ratio between the signal fluctuation amplitude and the time interval, and combining the signal feature parameters, the dynamic change rate of each signal is extracted to generate the change rate data; By monitoring the time series data of vibration signals and temperature signals, the fluctuation values ​​and average values ​​of each signal in different time intervals are extracted, and the change rate is obtained by calculating the ratio between the fluctuation amplitude and the time interval. Specifically, the vibration signal fluctuation values ​​within a certain period of time are collected. And temperature signal fluctuation value , using the formula: , calculate the rate of change data, where and Represents the time value of two time points respectively. and They represent the fluctuation values ​​corresponding to the vibration signals respectively. By calculating the change rate for each time interval, a complete change rate series data is generated. The change rate series is used for time series trend analysis. The change rate is classified in combination with the characteristics of the vibration signal, and the change characteristics of different categories of signals are extracted to generate change rate data.

[0037] Call the change rate data, establish a dynamic trend matrix, calculate the change amplitude and distribution law of the signal at the differentiated time nodes, and combine the signal fluctuation characteristics to extract the signal change range data at the time node; The change rate sequence is classified according to the time nodes to calculate the average change in each time period, and the signal change range is calculated according to the distribution law of the time node signal. Specifically, the time series signal is first divided into time nodes, and the formula is: , calculate the average rate of change of the time node signal, where Indicates time period The average rate of change within Indicates The rate of change of the signal, is the number of signals in the time period, and the average rate of change is compared with the fluctuation range of the signal within the time node, through the formula: , calculate the signal variation range, For time period The signal variation range within Represents all signal values ​​within the time period. Through the above process, a dynamic trend matrix is ​​established to generate the signal change range data at the time node.

[0038] Using the signal change range data of time nodes, a time distribution prediction model is established. Combining the dynamic changes of the signal with the time node distribution, the formula is adopted: ; Calculate and generate abnormal trend prediction values; in, represents the abnormal trend prediction value, Representative The rate of change of the signal, Representative The time node value of the class signal, represents the mean value of the time node, Represents the standard deviation of the time node, Representative The strength value of the signal, Represents the number of signal categories, Represents the number of time nodes, , They are the indexes of signal category and time node respectively.

[0039] formula: ; The benefit of the formula is that by combining the dynamic change rate of the signal and the Gaussian distribution model, it can achieve accurate prediction of the signal behavior at the time node, and enhance the sensitivity and adaptability of the model to abnormal trends; Detailed explanation of the formula and the process of formula calculation and derivation: Assume that the monitored signal data is , , , the corresponding time node , , , the mean value of the time node is , standard deviation The signal strengths are , , , then the calculation formula is derived as follows: ; ; ; ; ; ; The result shows that the calculated abnormal trend prediction value is 0.6329, which means that the abnormal trend of the signal at the current time node is close to the medium range, which can be used as a reference input for dynamic trend prediction and further used for the classification and response processing of signal anomaly prediction.

[0040] 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. The intelligent inspection system for power equipment based on artificial intelligence is characterized by: The system comprises: The fault mode generation module obtains the operating parameters of the power equipment through the inspection robot, inputs the current value, temperature value and vibration signal value, counts the amplitude distribution of the signal, extracts the frequency domain characteristics and change rate of the signal, calculates the dynamic characteristic range of the signal, establishes the fault characteristic mapping model, adjusts the parameters to match the signal distribution deviation, and generates the fault mode distribution value; The fault detection module obtains the current signal offset value and the frequency domain deviation in the inspection signal based on the fault mode distribution value, compares the interval difference between the inspection signal and the normal operating parameters, calculates the amplitude difference rate of the signal abnormality, establishes a fault feature deviation model, and generates the abnormal signal amplitude ratio; The fault classification module analyzes the characteristic values ​​of the current signal and the temperature signal in the differentiation interval based on the abnormal signal amplitude ratio, compares the overlap between the signal characteristic values ​​and the fault category interval, establishes a classification feature matrix, screens the signal contribution rate, and generates a fault type distribution coefficient; The inspection feature optimization module extracts the local trend value and global change rate of the current and temperature signals based on the fault type distribution coefficient, normalizes the signal category feature weight, dynamically matches the signal trend, extracts the inspection parameter matrix, and calculates the inspection feature weight distribution value; The abnormal prediction evaluation module analyzes the change rates of vibration signals and temperature signals based on the inspection feature weight distribution values, establishes a dynamic trend matrix, calculates the signal change range of time nodes, establishes a time distribution prediction model, and generates an abnormal trend prediction.

2. The artificial intelligence-based intelligent inspection system for electric power equipment according to claim 1 is characterized in that: The fault mode distribution value is specifically the signal amplitude distribution, signal frequency domain characteristics and change rate, and signal dynamic feature range. The fault feature deviation model includes the current signal offset value, frequency domain deviation, and signal abnormal amplitude difference rate. The abnormal signal amplitude ratio includes the current signal characteristic value, the temperature signal characteristic value, and the overlap between the signal characteristic value and the fault category interval. The fault type distribution coefficient is specifically the classification feature matrix and the signal contribution rate. The inspection feature weight distribution value is specifically the local trend value, the global change rate, and the signal category feature weight. The abnormal trend prediction value includes the vibration signal change rate, the temperature signal change rate, and the time node signal change range.

3. The power equipment intelligent inspection system based on artificial intelligence according to claim 2 is characterized in that: The steps for obtaining the failure mode distribution value are specifically as follows: Extract the amplitude distribution of current value, temperature value and vibration signal value, calculate the change rate of multiple signal amplitudes in the frequency domain, and generate amplitude distribution change rate parameters; Call the amplitude distribution change rate parameter, compare its value to see if it exceeds the dynamic feature range according to the upper and lower limits, classify the signal amplitude according to the frequency domain characteristics, extract the signal feature change value that meets the conditions of the upper and lower limits of the dynamic feature range, and calculate and generate the signal change rate within the dynamic feature range; The signal change rate within the dynamic feature range is called, the distribution deviation is analyzed, and the signal characteristic value mapping relationship is corrected based on the adjustment parameters, using the formula: ； Calculate and generate failure mode distribution values; in, represents the failure mode distribution value, Represents the rate of change of the signal characteristic value, represents the weight of the signal amplitude, Represents the lower limit of the dynamic feature range, Represents the upper limit of the dynamic feature range, represents the weight coefficient of the adjustment parameter, represents the signal feature mapping value, represents the mean value of signal characteristics, Represents the number of signal feature points, Represents the number of signal mapping points, , They are the indexes of signal feature points and mapping points respectively.

4. The artificial intelligence-based intelligent inspection system for electric power equipment according to claim 3 is characterized in that: The steps for obtaining the abnormal signal amplitude ratio are specifically as follows: According to the fault mode distribution value, the offset value and frequency domain deviation of the current signal in the inspection signal are obtained, the frequency domain deviation is used to statistically calculate the variation range of the differentiated signal, the offset value and the variation range are classified and sorted, and the current signal deviation data is generated; Calling the current signal deviation data, comparing the interval difference between the inspection signal and the normal operating parameters, calculating the deviation amplitude of the inspection signal in multiple intervals, analyzing the variation trend of the deviation amplitude, and counting the amplitude difference rate of the signal abnormality to generate signal abnormality amplitude difference data; Using the signal abnormal amplitude difference data, a fault feature deviation model is established using the formula: ； Generate abnormal signal amplitude ratio; in, represents the overall proportion of signal deviation, Representative The amplitude of the signal observation, Representative The normal amplitude reference of a signal, represents the number of all signal data points, is the index of the signal data point.

5. The artificial intelligence-based intelligent inspection system for electric power equipment according to claim 4 is characterized in that: The steps for obtaining the fault type distribution coefficient are specifically as follows: Based on the abnormal signal amplitude ratio, the characteristic values ​​of the current signal and the temperature signal in the differentiated interval are analyzed, and the fluctuation values ​​and frequency domain characteristics of the signals in the differentiated interval are classified item by item to generate differentiated characteristic data; The differentiated feature data is called, the overlap between the signal feature value and the fault category interval is compared, and the criticality of the multi-category interval is analyzed by counting and sorting the contribution rate of the signal amplitude in the overlapping interval and combining the signal feature weight parameter to generate the signal contribution rate data; Using the signal contribution rate data, a classification feature matrix is ​​established, and the weight relationship between the feature distribution of the differentiated signal category and the fault category interval is calculated using the formula: ； Calculate and generate the fault type distribution coefficient; in, represents the fault type distribution coefficient, Representative Class signal contribution rate, Representative The weight of the class signal, Representative The normal baseline contribution rate of the class signal, Representative Class The amplitude contribution of each interval is Represents the number of signal categories, represents the number of intervals, is the signal category index, is the interval index.

6. The artificial intelligence-based intelligent inspection system for electric power equipment according to claim 5 is characterized in that: The steps for obtaining the inspection feature weight distribution value are specifically as follows: Based on the fault type distribution coefficient, extract the local trend value and global change rate of the current and temperature signals, and generate local and global signal analysis data by analyzing the fluctuation characteristics of the signals in the local and global ranges item by item and combining the change law of the signal trend; Calling the local and global signal analysis data, normalizing the signal category feature weights, normalizing the signal categories to a unified range by calculating the ratio of multi-category signal feature values, and allocating feature weights of multi-category signals to generate normalized signal weight data; Using the normalized signal weight data, dynamically matching the signal trend, and extracting the inspection parameter matrix, the formula is: ； Calculate and generate inspection feature weight distribution values; in, Represents the inspection feature weight distribution value, Representative The local trend value of the class signal, Representative The global rate of change of the class signal, Representative The normalized weight of the class signal, Representative Stability index of class signal, Represents the number of signal categories, Signal category index.

7. The artificial intelligence-based intelligent inspection system for electric power equipment according to claim 6 is characterized in that: The steps for obtaining the abnormal trend prediction value are specifically as follows: Based on the inspection feature weight distribution value, the change rate of the vibration signal and the temperature signal is analyzed, and the dynamic change rate of each signal is extracted by calculating the change ratio between the signal fluctuation amplitude and the time interval, and combining the signal feature parameters to generate the change rate data; Call the change rate data, establish a dynamic trend matrix, calculate the change amplitude and distribution law of the signal at the differentiated time nodes, and extract the signal change range data at the time nodes in combination with the signal fluctuation characteristics; Using the signal change range data of the time nodes, a time distribution prediction model is established, combining the dynamic change of the signal with the time node distribution, and using the formula: ； Calculate and generate abnormal trend prediction values; in, represents the abnormal trend prediction value, Representative The rate of change of the signal, Representative The time node value of the class signal, represents the mean value of the time node, Represents the standard deviation of the time node, Representative The strength value of the signal, Represents the number of signal categories, Represents the number of time nodes, , They are the indexes of signal category and time node respectively.

Images