Fault prediction method and system based on electromechanical equipment operation information

Abstract

The invention provides a fault prediction method and system based on operation information of electromechanical equipment, and the method comprises the steps: collecting the operation information and operation condition information of the electromechanical equipment, and the environment interference information of the electromechanical equipment, and carrying out the data integration, and obtaining a comprehensive information set; performing feature importance evaluation based on information entropy on the comprehensive information set to obtain a feature importance sequence; based on the feature importance sequence, adaptively adjusting key parameters of a fault prediction algorithm based on a genetic algorithm to obtain an optimized fault prediction algorithm; inputting the operation information of the electromechanical equipment into the optimized fault prediction algorithm for fault prediction to obtain a fault prediction result; wherein the fault prediction result comprises the fault occurrence probability, the fault type and the fault occurrence position of the electromechanical equipment. According to the method, the defect that the current fault prediction algorithm lacks adaptive adjustment for the operation characteristics, the working condition change and the environment difference of the electromechanical equipment is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of electromechanical equipment, and in particular to a method and system for predicting faults based on electromechanical equipment operation information. Background Art

[0002] In the modern industrial field, electromechanical equipment is widely used in various key links such as production and manufacturing, energy supply, and infrastructure operation. The stable operation of electromechanical equipment plays a decisive role in ensuring production efficiency, product quality, and the overall safety and reliability of industrial systems.

[0003] Traditional maintenance methods for electromechanical equipment are mostly based on regular maintenance. Regular maintenance has many disadvantages. Its maintenance cycle is often set based on experience, which is difficult to accurately match the actual health status of the equipment. It will lead to unnecessary shutdown maintenance when the equipment is still in good operating condition, resulting in production interruption and waste of resources; or when the equipment has potential fault hazards but has not yet reached the maintenance cycle, it is not detected in time, which makes the fault worse, causing more serious equipment damage, production accidents and even casualties, accompanied by high maintenance costs and long downtime losses.

[0004] At the same time, existing fault prediction technologies mostly predict faults based on the operating data of electromechanical equipment. However, the parameter settings of most fault prediction algorithms are fixed, and they lack adaptability to the operating characteristics, operating conditions and environmental differences of different electromechanical equipment. They cannot flexibly adjust the prediction model according to actual conditions, which further limits the accuracy and generalization ability of fault prediction. Summary of the invention

[0005] The main purpose of the present invention is to provide a method and system for fault prediction based on electromechanical equipment operation information, aiming to overcome the defect that the current fault prediction algorithm lacks adaptive adjustment to the operating characteristics, working condition changes and environmental differences of electromechanical equipment.

[0006] To achieve the above object, the present invention provides a method for fault prediction based on electromechanical equipment operation information, comprising the following steps:

[0007] Collect the operation information, operation condition information and environmental interference information of electromechanical equipment and integrate the data to obtain a comprehensive information set;

[0008] Performing feature importance evaluation on the comprehensive information set based on information entropy to obtain a feature importance sequence;

[0009] Based on the feature importance sequence, key parameters of the fault prediction algorithm based on the genetic algorithm are adaptively adjusted to obtain an optimized fault prediction algorithm;

[0010] The operation information of the electromechanical equipment is input into the optimized fault prediction algorithm to perform fault prediction to obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the fault location.

[0011] Furthermore, the data integration includes: classifying and integrating the electromechanical equipment operation information collected from different sensors, the environmental interference information reflecting the environmental conditions of the equipment, and the operating condition information characterizing the current operating mode of the electromechanical equipment according to the data type and source to form a comprehensive information set;

[0012] Furthermore, the comprehensive information set is evaluated for feature importance based on information entropy to obtain a feature importance sequence, including:

[0013] Calculate the information entropy of each information in the comprehensive information set;

[0014] The degree of uncertainty of each piece of information is determined according to the size of the information entropy, and the feature importance sequence is obtained according to the degree of uncertainty.

[0015] Furthermore, the adaptive adjustment of key parameters of the fault prediction algorithm based on the genetic algorithm includes:

[0016] The key parameters in the fault prediction algorithm are encoded as chromosomes in the genetic algorithm, and the feature importance sequence is used as the input of the fitness function in the genetic algorithm. The key parameters on the chromosome are continuously optimized through the selection, crossover and mutation operations of the genetic algorithm until the preset convergence conditions are met.

[0017] Furthermore, the operation information of the electromechanical equipment is input into the optimized fault prediction algorithm to perform fault prediction, and a fault prediction result is obtained, including:

[0018] The operation information is divided into a plurality of information segments with time series characteristics, and a feature reconstruction process based on singular value decomposition is performed on each information segment to obtain a reconstructed feature matrix;

[0019] Each reconstructed feature matrix is ​​input into the support vector machine classifier in the optimized fault prediction algorithm, and the reconstructed feature matrix is ​​mapped to a high-dimensional space using the kernel function of the support vector machine classifier. The optimal classification hyperplane is found in the high-dimensional space, and the fault classification result corresponding to each information fragment is calculated;

[0020] Based on the dynamic time warping algorithm, the fault classification results of each information segment are integrated and corrected in time series to obtain the final fault prediction result; wherein, the dynamic time warping algorithm calculates the optimal matching path between the fault classification result sequences of different information segments, adjusts the classification result deviation caused by time difference, and thus determines the fault prediction result of the electromechanical equipment on the overall operation timeline.

[0021] Further, after obtaining the fault prediction result, the following steps are included:

[0022] Adding the comprehensive information set, the adjusted key parameters, and the characters in the fault prediction results to a preset graph structure template; the graph structure template includes a plurality of nodes, and adjacent nodes are connected by edges;

[0023] According to the characters on the adjacent nodes, the values ​​on the edges connecting the adjacent nodes are determined to obtain the character graph structure;

[0024] generating a communication key based on the character graph structure;

[0025] The comprehensive information set, the adjusted key parameters, and the fault prediction results are encrypted based on the communication key and sent to the management terminal.

[0026] Furthermore, the comprehensive information set, the adjusted key parameters, and the characters in the fault prediction results are added to a preset graph structure template, including:

[0027] Respectively obtaining the ratio of the number of characters in the comprehensive information set, the adjusted key parameters, and the fault prediction result;

[0028] According to the ratio, the preset graph structure template is divided into regions to obtain a first region, a second region, and a third region;

[0029] The characters in the comprehensive information set are added to each node in the first area, the characters in the adjusted key parameters are added to each node in the second area, and the characters in the fault prediction results are added to each node in the third area.

[0030] Further, generating a communication key based on the character graph structure includes:

[0031] Respectively obtaining the comprehensive information set, the adjusted key parameters, and the numeric characters in the fault prediction results;

[0032] Add the acquired digital characters to the matrix in sequence to generate a digital matrix; perform XOR calculation on the digital matrix to obtain an XOR matrix; the XOR matrix only includes the first number and the second number;

[0033] According to a preset rule, superimposing the XOR matrix into the character graph structure;

[0034] Obtain two farthest first numbers in the XOR matrix and connect them to obtain a first connection line; obtain two farthest second numbers in the XOR matrix and connect them to obtain a second connection line;

[0035] Searching for an edge intersecting the first connecting line and / or the second connecting line in the character graph structure as a target edge;

[0036] The assignments on the target edges are combined to obtain the communication key.

[0037] The present invention also provides a system for predicting faults based on electromechanical equipment operation information, comprising:

[0038] The acquisition module is used to collect the operation information, operation condition information and environmental interference information of the electromechanical equipment and integrate the data to obtain a comprehensive information set;

[0039] An evaluation module, used for performing feature importance evaluation on the comprehensive information set based on information entropy to obtain a feature importance sequence;

[0040] An adjustment module, used for adaptively adjusting key parameters of a fault prediction algorithm based on a genetic algorithm based on the feature importance sequence to obtain an optimized fault prediction algorithm;

[0041] The prediction module is used to input the operation information of the electromechanical equipment into the optimized fault prediction algorithm to perform fault prediction and obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the fault location.

[0042] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0043] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0044] The method and system for fault prediction based on the operation information of electromechanical equipment provided by the present invention include: collecting the operation information, operation condition information and environmental interference information of electromechanical equipment and integrating the data to obtain a comprehensive information set; performing feature importance evaluation based on information entropy on the comprehensive information set to obtain a feature importance sequence; based on the feature importance sequence, adaptively adjusting the key parameters of the fault prediction algorithm based on the genetic algorithm to obtain an optimized fault prediction algorithm; inputting the operation information of the electromechanical equipment into the optimized fault prediction algorithm to perform fault prediction and obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the location of the failure. In the present invention, based on the acquired operation information, operation condition information and environmental interference information of the electromechanical equipment, the key parameters of the fault prediction algorithm based on the genetic algorithm are adaptively adjusted so that it can adapt to the current application scenario. The defect that the current fault prediction algorithm lacks adaptive adjustment for the operation characteristics, working condition changes and environmental differences of electromechanical equipment is overcome. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a schematic diagram of the steps of a method for fault prediction based on electromechanical equipment operation information in one embodiment of the present invention;

[0046] Figure 2 is a block diagram of a system structure for performing fault prediction based on electromechanical equipment operation information in one embodiment of the present invention;

[0047] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0048] The implementation, functional features and advantages of the present invention will be further described in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0049] 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.

[0050] Reference Figure 1 In one embodiment of the present invention, a method for fault prediction based on electromechanical equipment operation information is provided, comprising the following steps:

[0051] Step S1, collecting operation information, operation condition information and environmental interference information of electromechanical equipment and integrating the data to obtain a comprehensive information set;

[0052] Step S2, performing feature importance evaluation on the comprehensive information set based on information entropy to obtain a feature importance sequence;

[0053] Step S3, based on the feature importance sequence, adaptively adjusting key parameters of the fault prediction algorithm based on the genetic algorithm to obtain an optimized fault prediction algorithm;

[0054] Step S4, inputting the operation information of the electromechanical equipment into the optimized fault prediction algorithm to perform fault prediction and obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the fault location.

[0055] In this embodiment, as described in step S1 above, operation information is obtained by installing various sensors at key parts of electromechanical equipment. For example, temperature sensors can monitor the working temperature of the equipment. They are distributed in motor windings, bearings and other parts that are prone to heat, and provide real-time feedback of temperature data, because excessive temperature indicates problems such as heat dissipation failure and increased wear of components; vibration sensors are installed on the housing or shaft system of the equipment to detect parameters such as vibration amplitude and frequency of the equipment during operation. Abnormal vibration is often related to equipment failure states such as imbalance, looseness, and resonance; current transformers and voltage transformers are used to collect electrical parameters of the equipment, such as the operating current and voltage of the motor. Excessive current indicates electrical faults such as motor overload and short circuit, and voltage fluctuations affect the normal operation stability and power output of the equipment. The above sensors transmit the collected analog signals or digital signals to the data acquisition system, and the data acquisition system collects and preliminarily processes the signals according to a certain sampling frequency, such as filtering to remove noise interference, amplifying weak signals, etc., to ensure that the collected operation information is accurate, reliable and representative.

[0056] Read operating condition information from the control system of electromechanical equipment. For example, the load condition of the equipment can be indirectly inferred from the load sensor data in the control system or from the output power, torque and other parameters of the motor. Different load levels have different effects on the wear, heat generation, energy consumption and other aspects of the equipment. The operating speed information can be directly obtained from the speed sensor of the equipment or the speed feedback module of the control system. The stress condition, vibration characteristics and lubrication requirements of the mechanical parts of the equipment are different at different operating speeds. The start-stop number record reflects the frequency of use and working cycle of the equipment. Frequent start-stop operations lead to premature fatigue damage of the electrical components and mechanical connectors of the equipment.

[0057] Use environmental monitoring sensors to collect environmental interference information. For example, temperature and humidity sensors are used to monitor the temperature and humidity of the environment in which the equipment is located. In some humidity-sensitive electronic equipment or electromechanical equipment with metal parts that are prone to rust and corrosion in high temperature and high humidity environments, environmental temperature and humidity data are crucial for fault prediction; electromagnetic interference sensors detect the electromagnetic intensity and frequency in the surrounding environment. In an environment with a large number of electromagnetic devices or close to strong electromagnetic sources such as high-voltage transmission lines, electromechanical equipment will be subject to electromagnetic interference, affecting the normal operation of its control system and the measurement accuracy of the sensor.

[0058] The collected operation information, operation condition information and environmental interference information are aligned according to the timestamp to ensure that the various types of data at the same time can be matched. Then, the above data are sorted and merged according to the preset data structure. For example, a multidimensional array or database table can be constructed, where each row represents a data record at a time point, and each column corresponds to a different type of information, such as temperature, vibration, load, ambient temperature, electromagnetic interference intensity, etc., to form a comprehensive information set.

[0059] As described in step S2 above: Information entropy is a measure of data uncertainty or information volume. For each feature in the comprehensive information set (such as temperature, vibration, etc.), first count the distribution of its values ​​in the entire data set. For example, for the temperature feature, determine the frequency of occurrence of data within different value ranges. The information entropy is calculated based on the frequency. The larger the information entropy, the more dispersed the value of the feature is, and the more information it contains; conversely, the smaller the information entropy, the more concentrated the feature value is, and the less effective information is provided.

[0060] After calculating the information entropy of each feature, the feature importance coefficient is determined by comparing it with the total information entropy of the comprehensive information set. The feature with a larger importance coefficient plays a more critical role in fault prediction. All features are sorted according to the calculated feature importance coefficients to form a feature importance sequence. For example, if the importance coefficient of the vibration feature is greater than the importance coefficient of the temperature feature, the vibration feature is ranked before the temperature feature in the sequence. This sequence will provide a basis for the adaptive adjustment of the key parameters of the subsequent fault prediction algorithm, so that the algorithm can pay more attention to important features and improve the accuracy of fault prediction.

[0061] As described in step S3 above, the genetic algorithm is an optimization algorithm based on biological evolution theory, which searches for the optimal solution by simulating genetic operations such as natural selection, crossover and mutation. In the fault prediction algorithm, the problem needs to be encoded first, and the parameters of the fault prediction model (such as the weights and thresholds of the neural network) are represented as chromosomes (a string of binary or real number codes). Then, the initial population is randomly generated, and each individual represents a set of possible parameter combinations.

[0062] Adjust the key parameters of the genetic algorithm according to the feature importance sequence. For the encoding method, give finer encoding accuracy to the parameters corresponding to the features with high importance. For example, if the vibration feature is determined to be very important, then more encoding bits are used when encoding the fault prediction model parameters related to the vibration feature so that these parameters can be adjusted more accurately during the search process of the genetic algorithm. When determining the crossover probability and mutation probability, for the parameters related to important features, reduce the mutation probability to maintain their stability during the evolution process and avoid losing the existing excellent characteristics due to excessive mutation; at the same time, appropriately adjust the crossover probability so that the information of the parameters related to important features can be better retained in the genetic operation, and promote the combination and propagation of excellent genes. For the parameters related to minor features, the mutation probability can be appropriately increased to increase the algorithm's exploration of these parameters and prevent the algorithm from falling into the local optimal solution.

[0063] Calculate the fitness value of each individual (parameter combination) in the population. The fitness function is usually defined based on the prediction error of the fault prediction model. The smaller the prediction error, the higher the fitness value. Through the selection operation, select excellent individuals according to the fitness ratio to enter the next generation population; then perform a crossover operation to exchange some gene fragments between individuals to generate new individuals; finally, perform a mutation operation to randomly change some gene bits of the individual and introduce new gene information. After multiple generations of genetic evolution, the parameters of the fault prediction algorithm are continuously adjusted, so that the prediction performance of the algorithm is gradually improved, and finally an optimized fault prediction algorithm is obtained. The algorithm can better adapt to the operating characteristics and failure modes of electromechanical equipment and improve the accuracy and reliability of fault prediction.

[0064] As described in step S4 above, before the operation information of the electromechanical equipment is input into the optimized fault prediction algorithm, it needs to be preprocessed. First, the data is cleaned to remove obvious outliers. For example, for temperature data, if the temperature value of a certain sampling point is far beyond the normal operating temperature range and does not match the temperature change trend of adjacent time points, it may be an outlier caused by sensor failure or data acquisition error, which should be removed. Then, the data is normalized to map feature data of different magnitudes to the same numerical range, such as the [0,1] interval. This can avoid affecting the performance of the fault prediction algorithm due to large differences in data magnitude, and improve the convergence speed and accuracy of the algorithm. The preprocessed operation information is input into the optimized fault prediction algorithm.

[0065] In one embodiment, if the fault prediction algorithm adopts a neural network model, the data will be calculated in turn through the input layer, hidden layer, and output layer. The input layer receives the normalized operation information data and passes it to the hidden layer; the neurons in the hidden layer perform nonlinear transformations on the input data through activation functions (such as ReLU functions) to extract complex features and patterns in the data; finally, the output layer calculates the fault prediction results based on the output results of the hidden layer, combined with the weights and thresholds obtained by pre-training. For example, the output layer has multiple neurons, one of which outputs the probability value of the failure of the electromechanical equipment, and the other neurons determine the fault type (such as motor failure, transmission system failure, etc.) and the location where the fault may occur (such as motor bearings, gearboxes, etc.) through classification.

[0066] Interpret and judge based on the output results of the fault prediction algorithm. If the probability of a fault exceeds the preset threshold (such as 0.8), it is considered that the equipment has a high risk of failure. At this time, the corresponding fault type and fault location information will provide targeted maintenance guidance for equipment maintenance personnel. For example, if the prediction results show that the probability of motor bearing failure is 0.9 and the fault type is bearing wear, then the maintenance personnel can focus on checking the lubrication condition and wear degree of the motor bearing, and take timely maintenance measures such as replacing bearings or adding lubricants, thereby realizing electromechanical equipment management based on predictive maintenance, improving equipment reliability, reducing maintenance costs and reducing the risk of production interruptions.

[0067] In one embodiment, the data integration includes: classifying and integrating the electromechanical equipment operation information collected from different sensors, the environmental interference information reflecting the environmental conditions of the equipment, and the operating condition information characterizing the current operating mode of the electromechanical equipment according to the data type and source to form a comprehensive information set;

[0068] In this embodiment, the received operation information sequence, environmental interference information sequence and operating condition information sequence are finally classified and integrated according to the data type and source. First, a multidimensional data structure is established. For example, it can be in the form of a three-dimensional array or a database table, in which one dimension represents the data type (such as operation information, environmental interference information, operating condition information), one dimension represents the data source (such as a specific sensor number or control system module), and the other dimension represents the time sequence. Then, the data in each sequence is filled into the data structure in sequence according to the corresponding dimension and time sequence to form a complete comprehensive information set. Such a comprehensive information set can comprehensively and systematically reflect the operating status, environment and operating conditions of electromechanical equipment at a specific point in time.

[0069] In one embodiment, the comprehensive information set is evaluated for feature importance based on information entropy to obtain a feature importance sequence, including:

[0070] Calculate the information entropy of each information in the comprehensive information set;

[0071] The degree of uncertainty of each piece of information is determined according to the size of the information entropy, and the feature importance sequence is obtained according to the degree of uncertainty.

[0072] In this embodiment, information entropy is used to measure a characteristic of each information in the comprehensive information set. It mainly reflects the uncertainty of the information, that is, the degree to which the value of the information is elusive, or a measure of how many unexpected situations it can bring. For example, if the value of a piece of information is always fixed or changes very little, then its uncertainty is low and the information entropy is also small.

[0073] To calculate the information entropy of each piece of information in the comprehensive information set, you must first determine the value of this information in the entire data set. For example, if you need to understand the information entropy of the temperature information in the operation information of electromechanical equipment, you need to sort out all the collected values ​​about temperature. Count the number of times each different temperature value or temperature value within a certain temperature range appears. For example, divide the temperature range into several small intervals, and then count how many times the temperature value in each interval appears in the collected data. According to the proportion of the number of times these different values ​​or value intervals appear in the total number of times, we can further determine the degree of uncertainty of this information, that is, calculate its information entropy. For other information in the comprehensive information set, such as vibration amplitude, ambient humidity, load size in operating conditions, etc., the same idea is used to count the number of times their different values ​​or value intervals appear, and then calculate their respective information entropy.

[0074] The size of information entropy is directly linked to the degree of uncertainty of each piece of information. If the information entropy of a piece of information is large, it means that the value of this information changes a lot, and it is difficult to guess in advance what value it will take next time, so its degree of uncertainty is very high. For example, the vibration frequency of electromechanical equipment during operation, if its information entropy is large, it means that in the collected data, the vibration frequency is sometimes high and sometimes low, and there is no fixed pattern. It is difficult to accurately predict the next vibration frequency based on previous data, so its degree of uncertainty is high.

[0075] According to the calculated information entropy of each information, its uncertainty can be determined. Then, the feature importance sequence is generated according to this uncertainty. Generally speaking, the information is arranged directly in order from large to small information entropy. The information with the largest information entropy means that it has the highest degree of uncertainty and contains the largest amount of information, so it is placed at the front of the feature importance sequence. Then, in the order of decreasing information entropy, the information is arranged in order, and the information with the smallest information entropy is placed at the end.

[0076] In one embodiment, the adaptive adjustment of key parameters of the fault prediction algorithm based on the genetic algorithm includes:

[0077] The key parameters in the fault prediction algorithm are encoded as chromosomes in the genetic algorithm, and the feature importance sequence is used as the input of the fitness function in the genetic algorithm. The key parameters on the chromosome are continuously optimized through the selection, crossover and mutation operations of the genetic algorithm until the preset convergence conditions are met.

[0078] In this embodiment, when the genetic algorithm is used to optimize the key parameters of the fault prediction algorithm, the first thing to do is to encode these key parameters into a form that the genetic algorithm can process, that is, a chromosome form. The main purpose of this step is to allow the genetic algorithm to process the parameter information of the fault prediction algorithm like processing gene information in biological inheritance, so that the optimal combination of these parameters can be searched through various operations of the genetic algorithm later.

[0079] Assume that there are some key parameters in the fault prediction algorithm, such as weights, thresholds and other parameters in the neural network model (the fault prediction algorithm uses a neural network architecture), or other important parameters that determine the performance of the algorithm. These parameters can be encoded into chromosomes using binary coding or real number coding. Taking binary coding as an example, if the value range of a key parameter is between [0,10], this range can be divided into several intervals, and then each interval can be represented by a binary number. For example, if [0,10] is divided into 100 equally spaced intervals, each interval can be represented by a 7-bit binary number. In this way, a key parameter is encoded into a string of binary numbers, and multiple key parameters are combined to form a chromosome, just like the genome in an organism synthesizes chromosomes. Through this encoding method, the genetic algorithm can operate and optimize these encoded chromosomes (that is, the information representing the combination of key parameters).

[0080] The fitness function mentioned above plays a vital role in the genetic algorithm. It is like a criterion to measure the quality of each individual (that is, the key parameter combination represented by each chromosome) in solving the problem (that is, optimizing the fault prediction algorithm to achieve more accurate fault prediction). The feature importance sequence is used as the input of the fitness function so that the fitness function can more accurately evaluate the quality of each key parameter combination based on the importance of the feature.

[0081] When the feature importance sequence is input into the fitness function, the fitness function will make a comprehensive evaluation based on the importance of each feature in the sequence and the effect of the key parameter combination represented by the current chromosome on fault prediction. For example, if the feature importance sequence indicates that a certain feature is very important for fault prediction, and the key parameter combination represented by the current chromosome enables the fault prediction algorithm to show good accuracy when processing data related to this important feature, then the evaluation value of the individual corresponding to this chromosome in the fitness function will be relatively high. Conversely, if the key parameter combination does not work well when processing data related to important features, then its fitness value will be low. In this way, by incorporating the feature importance sequence into the scope of consideration of the fitness function, the genetic algorithm can be guided to search for the optimal key parameter combination in a direction that is more conducive to using important features for accurate fault prediction.

[0082] The main function of the selection operation is to select relatively excellent individuals from the current population (that is, a set of multiple chromosomes, each chromosome represents a key parameter combination) so that these excellent individuals have more opportunities to pass on their "excellent genes" (that is, better key parameter combinations) to the next generation. When performing the selection operation, the probability of selection is usually determined based on the evaluation value of the individual in the fitness function. The higher the evaluation value of the individual, the greater the probability of being selected. For example, the roulette selection method can be used to normalize the fitness values ​​of all individuals in the population so that their sum is 1. Then, the area ratio occupied by each individual on a virtual roulette wheel is determined based on the normalized fitness value of each individual. When making a selection, it is equivalent to randomly rotating the pointer on this roulette wheel. The individual is selected within the area occupied by the individual where the pointer stops. Through such a selection operation, those key parameter combinations (that is, individuals with high fitness values) that enable the fault prediction algorithm to perform better when processing data related to important features have a greater chance of being selected into the next generation population, thereby providing a better foundation for subsequent optimization operations.

[0083] The crossover operation imitates the gene crossover phenomenon in the biological genetic process. Its purpose is to introduce new gene combinations into the population and increase the diversity of the population so that it is possible to find a better combination of key parameters. When performing the crossover operation, two chromosomes (that is, two key parameter combinations) are first randomly selected from the population obtained after the selection operation. Then, according to a certain crossover method, such as single-point crossover, multi-point crossover or uniform crossover, some gene fragments of the two chromosomes are exchanged to form two new chromosomes. For example, in a single-point crossover, a point is randomly selected on the two chromosomes, and then the gene fragments after this point are exchanged, so that two new chromosomes are obtained. Through the crossover operation, the originally different key parameter combinations can be integrated with each other to produce a better combination, which helps to improve the performance of the fault prediction algorithm.

[0084] The mutation operation simulates the gene mutation phenomenon in the biological genetic process. Its main function is to introduce some random changes in the population to prevent the population from converging to the local optimal solution too early, but to continue to explore a wider search space and find a better combination of key parameters. When performing the mutation operation, each chromosome (that is, each key parameter combination) in the population after the selection and crossover operations will be randomly changed. For example, for chromosomes using binary coding, some bits can be randomly selected, and then the values ​​of these bits can be reversed (from 0 to 1 or from 1 to 0); for chromosomes using real number coding, the values ​​of certain key parameters can be randomly increased or decreased within a preset range. Through the mutation operation, even if the population has a certain optimization direction after the selection and crossover operations, it can continue to explore other key parameter combinations through these random changes to improve the performance of the fault prediction algorithm.

[0085] The preset convergence condition is used to determine whether the genetic algorithm has reached a relatively stable optimization state, that is, whether a relatively satisfactory key parameter combination has been found so that the performance of the fault prediction algorithm reaches an acceptable level. The convergence condition can be set in many ways, such as setting the average fitness value of the population to change within several consecutive generations less than a certain threshold, or setting the fitness value of the best individual in the population to have no obvious improvement within several consecutive generations. When these convergence conditions are met, it means that the genetic algorithm has fully optimized the key parameters on the chromosome through continuous selection, crossover, and mutation operations, and the key parameter combination obtained can enable the fault prediction algorithm to achieve better results in dealing with the fault prediction problem of electromechanical equipment. At this time, the operation of the genetic algorithm can be stopped, and the optimized key parameter combination can be applied to the fault prediction algorithm to obtain the optimized fault prediction algorithm.

[0086] In one embodiment, the operation information of the electromechanical equipment is input into the optimized fault prediction algorithm to perform fault prediction, and a fault prediction result is obtained, including:

[0087] The operation information is divided into a plurality of information segments with time series characteristics, and a feature reconstruction process based on singular value decomposition is performed on each information segment to obtain a reconstructed feature matrix;

[0088] Each reconstructed feature matrix is ​​input into the support vector machine classifier in the optimized fault prediction algorithm, and the reconstructed feature matrix is ​​mapped to a high-dimensional space using the kernel function of the support vector machine classifier. The optimal classification hyperplane is found in the high-dimensional space, and the fault classification result corresponding to each information fragment is calculated;

[0089] Based on the dynamic time warping algorithm, the fault classification results of each information segment are integrated and corrected in time series to obtain the final fault prediction result; wherein, the dynamic time warping algorithm calculates the optimal matching path between the fault classification result sequences of different information segments, adjusts the classification result deviation caused by time difference, and thus determines the fault prediction result of the electromechanical equipment on the overall operation timeline.

[0090] In this embodiment, the operation information of the electromechanical equipment is usually a data sequence that changes continuously over time. The main purpose of dividing it into multiple information segments with time series characteristics is to more carefully analyze the characteristic changes of the operation information in different time periods, so as to better capture the local characteristics related to the fault. This segmentation can be performed at fixed time intervals, such as dividing the operation information into independent information segments at certain sampling time intervals (such as every hour, every minute, etc., and the specific interval can be determined according to the equipment operation characteristics and data collection frequency). After such division, each information segment represents the operation status information of the equipment in a specific time period, and retains the characteristics of the time series, that is, the order of the data reflects the evolution of the equipment operation status over time.

[0091] Singular value decomposition (SVD) is an important matrix decomposition method that plays a key role in feature processing of each information fragment. For each information fragment, it can be regarded as data in the form of a matrix (for example, if the information fragment contains multiple different types of operating parameters, such as temperature, vibration, etc., the values ​​of these parameters at different time points can be arranged into a matrix). Through singular value decomposition, this matrix can be decomposed into the product form of three matrices. Based on the results of singular value decomposition, feature reconstruction processing can be performed. Usually, the matrix elements corresponding to the part with larger singular values ​​are selected to reconstruct the feature matrix, because the size of the singular value reflects the importance of the corresponding feature, and the features corresponding to larger singular values ​​often contain more key information about the operating status of the equipment. The reconstructed feature matrix reconstructed in this way can highlight the feature information that is more important for fault prediction in each information fragment, reduce the interference of noise and redundant information in the original data, and provide a more representative and effective data basis for subsequent fault prediction.

[0092] The support vector machine (SVM) classifier is a very effective machine learning classification algorithm, which is widely used in the field of fault prediction. Its core idea is to divide different categories of data as clearly as possible by finding an optimal classification hyperplane. In this technical solution, the reconstructed feature matrix is ​​input into the support vector machine classifier to use its powerful classification ability to distinguish different fault states and normal states in the operation information of electromechanical equipment. The support vector machine classifier has the following advantages: it can show good performance when processing high-dimensional data, and even when the data samples are relatively small, it can achieve more accurate classification by reasonably selecting kernel functions and other methods. Moreover, it has a certain resistance to local disturbances of the data, and can avoid large changes in classification results due to small fluctuations in the data to a certain extent.

[0093] In practical applications, the reconstructed feature matrix obtained after feature reconstruction of the operation information of electromechanical equipment is still in a relatively low-dimensional space, but sometimes the data cannot be effectively classified by a simple linear classification hyperplane in the low-dimensional space. At this time, the kernel function of the support vector machine classifier plays an important role. The kernel function can map the data points in the reconstructed feature matrix to a higher-dimensional space, so that the data that seems to be nonlinearly separable in the low-dimensional space becomes linearly separable in the high-dimensional space. Common kernel functions include linear kernels, polynomial kernels, Gaussian kernels, etc. By selecting a suitable kernel function and mapping the reconstructed feature matrix to a high-dimensional space, it is easier to find the optimal classification hyperplane in this new high-dimensional space. This optimal classification hyperplane can separate the reconstructed feature matrix data of different fault states and normal states with the lowest misclassification rate, thereby providing a basis for calculating the fault classification results corresponding to each information fragment.

[0094] Once the optimal classification hyperplane is found in the high-dimensional space, the corresponding fault classification results can be calculated based on the positional relationship of each reconstructed feature matrix data point relative to the hyperplane. Generally speaking, if a reconstructed feature matrix data point is located on one side of the hyperplane, it can be judged to belong to a certain type of fault state; if it is located on the other side of the hyperplane, it is judged to belong to another type of fault state (for example, normal state or other different types of fault states). By making such a judgment on each reconstructed feature matrix, the fault classification results corresponding to each information fragment can be obtained. These results preliminarily reflect the preliminary judgment of the fault possibility and fault type of the equipment in each time period (that is, the time period corresponding to each information fragment).

[0095] In the previous steps, the fault classification results corresponding to each information fragment have been obtained, but these results are obtained based on the independent analysis of each information fragment, and do not fully consider the temporal coherence between different information fragments and the deviation of classification results caused by time differences. For example, the fault classification results of the equipment in a certain time period are affected by the changes in the equipment operating status in the previous time period, or due to factors such as different data collection time intervals and dynamic changes in equipment operating conditions, the fault classification results between adjacent information fragments are incoherent or inconsistent. The dynamic time warping algorithm (DTW) is introduced to solve this problem. Its basic principle is to find a way to make these result sequences match as much as possible in time by calculating the optimal matching path between the fault classification result sequences of different information fragments, thereby adjusting the deviation of classification results caused by time differences.

[0096] When the fault classification results of each information fragment are input into the dynamic time warping algorithm, the algorithm first constructs a distance matrix, the elements of which represent a certain distance measure (such as Euclidean distance, etc.) between the fault classification result sequences of different information fragments. Then, the optimal matching path is found in this distance matrix through a dynamic search algorithm (such as an algorithm based on dynamic programming). This path represents a combination method that can achieve the best matching of the fault classification result sequences of each information fragment in time. Once the optimal matching path is found, the fault classification results of each information fragment can be integrated and corrected in time according to this path. Specifically, along this optimal matching path, the fault classification results of different information fragments are weighted averaged or otherwise integrated in a reasonable manner, and the existing classification result deviations are corrected according to the differences in the path. Through such time integration and correction processing, a more coherent and accurate fault prediction result on the overall operation time axis can be obtained. This result can better reflect the fault possibility, fault type, and approximate time range of the fault of the electromechanical equipment during the entire operation process, and provide equipment maintenance personnel with a more comprehensive and accurate equipment fault prediction situation so that corresponding maintenance measures can be taken in time.

[0097] In one embodiment, after obtaining the fault prediction result, the following steps are included:

[0098] Adding the comprehensive information set, the adjusted key parameters, and the characters in the fault prediction results to a preset graph structure template; the graph structure template includes a plurality of nodes, and adjacent nodes are connected by edges;

[0099] According to the characters on the adjacent nodes, the values ​​on the edges connecting the adjacent nodes are determined to obtain the character graph structure;

[0100] generating a communication key based on the character graph structure;

[0101] The comprehensive information set, the adjusted key parameters, and the fault prediction results are encrypted based on the communication key and sent to the management terminal.

[0102] In this embodiment, the preset graph structure template is used as a framework for data organization and association. It is designed to integrate the comprehensive information set, the adjusted key parameters, and the relevant character information in the fault prediction results in a structured and logical manner. This graph structure template contains multiple nodes, each of which can be regarded as an information storage unit for storing specific characters extracted from the above-mentioned types of data. Adjacent nodes are connected by edges, which not only play a physical role in connecting nodes, but more importantly, they will provide a basis for the subsequent determination of the assignment on the edges and the construction of the relationship between characters.

[0103] The assignment of values ​​to the edges connecting adjacent nodes is based on a certain association relationship between the characters placed on the adjacent nodes. This association relationship can be defined and calculated in a variety of ways, depending on the characteristics of the data being processed and the encryption or information association effect you want to achieve.

[0104] A common approach is based on the numerical attributes of characters (if characters can be converted into numerical form, such as digital characters or characters that can be converted into numerical values ​​through encoding). For example, if the temperature value character and the corresponding vibration frequency value character of the electromechanical equipment at a certain moment are placed on adjacent nodes, the difference, ratio or some custom mathematical operation result of the two values ​​can be calculated as the assignment of the edge connecting the two nodes. Another approach is based on the logical relationship of characters. For example, if a node is placed with the fault type character in the fault prediction result, and another adjacent node is placed with the operating condition character related to the fault type in the comprehensive information set, then the edge assignment can be determined based on whether there is a specific logical association (such as the fault type is "overheating" and the operating condition is "high load and long-term operation", then a corresponding logical value can be assigned to the edge according to the degree of this causal association). In this way, the edges between each pair of adjacent nodes are assigned according to different association rules, thereby constructing a complete character graph structure. This character graph structure not only contains the character information of the original data, but also reflects the internal relationship between the characters through the assignment of edges.

[0105] The process of generating communication keys based on the character graph structure utilizes the information contained in the nodes and edges in the character graph structure and the complex relationships between them. One generation mechanism is to obtain a unique string with a specific length and format by performing some mathematical transformation or encoding operation on the character graph structure, which will be used as the communication key.

[0106] For example, all node characters and edge values ​​in the character graph structure can be arranged and combined in a certain order (for example, first in the order of node numbering, then in the order of edge connection) to form a long string. Then a hash operation is performed on this long string to obtain a hash value of a fixed length. This hash value can be used as the basis for the communication key, and its format can be further adjusted as needed (such as truncating a specific length, adding a specific prefix or suffix, etc.), and finally a communication key that meets the communication encryption requirements is generated. The uniqueness and randomness of the communication key generated in this way come from the complexity of the character graph structure and the mathematical operations used, which can ensure the security and confidentiality of the communication to a certain extent.

[0107] The purpose of using the generated communication key to encrypt the comprehensive information set, adjusted key parameters, and fault prediction results is to ensure the security and confidentiality of these important data during transmission.

[0108] In one embodiment, the characters in the comprehensive information set, the adjusted key parameters, and the fault prediction results are added to a preset graph structure template, including:

[0109] Respectively obtaining the ratio of the number of characters in the comprehensive information set, the adjusted key parameters, and the fault prediction result;

[0110] According to the ratio, the preset graph structure template is divided into regions to obtain a first region, a second region, and a third region;

[0111] The characters in the comprehensive information set are added to each node in the first area, the characters in the adjusted key parameters are added to each node in the second area, and the characters in the fault prediction results are added to each node in the third area.

[0112] In this embodiment, first, the ratio of the number of characters in the comprehensive information set, the adjusted key parameters, and the fault prediction results is obtained. The main purpose is to reasonably divide the preset graph structure template according to the relative size of their respective data volumes, so that each part has a suitable area in the graph structure for character placement, thereby achieving balanced distribution and effective organization of data in the graph structure. This division method based on the ratio of data volume helps to present the relationship between the various parts of data more clearly in the future, and can fully consider the characteristics and weights of different parts of data in subsequent operations such as generating communication keys.

[0113] First, it is necessary to determine how to count the number of characters in each part. For the comprehensive information set, it contains data on various aspects such as electromechanical equipment operation information, environmental interference information, and operating condition information, and these data exist in various formats (such as numbers, letters, symbols, etc.). For example, if the comprehensive information set is stored in the form of a database table, then each row and column in the table can be traversed, and the characters (including numeric characters, text description characters, etc.) in it can be counted to obtain the total number of characters in the comprehensive information set. Similarly, for the adjusted key parameters, which are some numerical parameters processed by the optimization algorithm, these parameters are presented in a certain standard character representation (such as scientific notation, decimal representation of a specific precision, etc.), and the number of characters in them is counted. For fault prediction results, such as text descriptions of fault types, numerical values ​​of fault probability, and identification of fault locations, the number of characters is also calculated according to the corresponding character statistics rules. Then, the number of characters in these three parts are compared in pairs, and the ratios between them are calculated, such as the ratio of the number of characters of the comprehensive information set to the adjusted key parameters, the ratio of the number of characters of the comprehensive information set to the fault prediction results, and the ratio of the number of characters of the adjusted key parameters to the fault prediction results.

[0114] The area of ​​the preset graph structure template is divided according to the ratio of the number of characters calculated above. The principle followed is to match the size of each area with the relative proportion of the number of characters in the corresponding data part. In other words, the part with a larger number of characters will be allocated to a relatively larger area in the graph structure template so that it can accommodate more character information; while the part with a smaller number of characters will be allocated to a relatively smaller area. This ensures that when characters are added later, each part of the data can be placed in order in its own appropriate area, and there will be no situation where one area is too crowded and another area is too empty, thereby achieving effective use of the graph structure template space and reasonable layout of data.

[0115] Assuming that the number of characters in the comprehensive information set is the largest compared to the number of characters in the adjusted key parameters and fault prediction results, and the ratio of the number of characters in the comprehensive information set to the adjusted key parameters is 3:1, and the ratio of the number of characters in the comprehensive information set to the adjusted key parameters is 4:1, then the graph structure template can be divided according to this ratio. For example, the first 60 nodes (approximately 60%) of the graph structure template can be divided into the first area for placing the characters of the comprehensive information set; the next 20 nodes (approximately 20%) are divided into the second area for placing the characters of the adjusted key parameters; the last 20 nodes (approximately 20%) are divided into the third area for placing the characters of the fault prediction results. Of course, the specific division method can be flexibly adjusted according to the actual shape of the graph structure template, node distribution, etc.

[0116] The characters in each part of the data are added to the corresponding regional nodes respectively, in order to further organize the data in a structured manner, so that the comprehensive information set, the adjusted key parameters and the fault prediction results each form a relatively independent and interrelated information set in the graph structure template. In this way, in subsequent processing (such as determining edge assignments, generating communication keys, etc.), it is more convenient to operate on data in different regions, and it can clearly show the relationship between each part of the data and their contribution to the overall fault prediction related information.

[0117] In one embodiment, generating a communication key based on the character graph structure includes:

[0118] Respectively obtaining the comprehensive information set, the adjusted key parameters, and the numeric characters in the fault prediction results;

[0119] Add the acquired digital characters to the matrix in sequence to generate a digital matrix; perform XOR calculation on the digital matrix to obtain an XOR matrix; the XOR matrix only includes the first number and the second number;

[0120] According to a preset rule, superimposing the XOR matrix into the character graph structure;

[0121] Obtain two farthest first numbers in the XOR matrix and connect them to obtain a first connection line; obtain two farthest second numbers in the XOR matrix and connect them to obtain a second connection line;

[0122] Searching for an edge intersecting the first connecting line and / or the second connecting line in the character graph structure as a target edge;

[0123] The assignments on the target edges are combined to obtain the communication key.

[0124] In this embodiment, digital characters are obtained from the comprehensive information set, the adjusted key parameters and the fault prediction results because the digital characters often carry key quantitative information in these data, such as the temperature value, current value and other specific values ​​of the electromechanical equipment in the comprehensive information set, the weight, threshold and other numerical values ​​in the algorithm in the adjusted key parameters, and the characters corresponding to the numerical values ​​such as the probability of fault occurrence in the fault prediction results. By extracting these digital characters, it is possible to focus on the key quantitative elements in the data, providing basic materials for the subsequent generation of communication keys with specific properties and security, so that the generated keys can be closely related to these important data features, thereby ensuring the relevance and adaptability of the data encryption and decryption process to the original data to a certain extent.

[0125] For the comprehensive information set, if it is stored in a certain data structure (such as a database table, array, etc.), it is necessary to traverse each record or element therein to determine whether each character is a numeric character. For example, for a table that records the operating temperature, vibration frequency, current and other information of electromechanical equipment, when encountering characters such as "35" (temperature value) and "120" (current value), they are extracted as numeric characters. When extracting characters, various symbols, decimal points, etc. should be removed. For the adjusted key parameters, assuming that the optimized neural network weight values ​​and other parameters are stored in a specific format, such as "0.356", "-1.2", etc., the numeric character parts (0, 3, 5, 6, 1, 2) are also extracted. For the fault prediction results, if the numerical values ​​such as the fault probability "0.8" exist in the form of characters, these numeric characters are also extracted. Through such comprehensive and detailed traversal and extraction operations, it is ensured that all relevant numeric characters in each part of the data are obtained.

[0126] The purpose of adding the acquired digital characters from different data sources to the matrix in sequence is to organize and process these digital characters in a structured manner. They can be added to the rows or columns of the matrix one by one in a certain order, such as according to the data source (the order of comprehensive information set, adjusted key parameters, and fault prediction results), and then in each source according to the order in which the digital characters appear in the original data, thus forming a digital matrix. This matrix form facilitates subsequent unified mathematical operations, integrating the originally scattered digital characters into a regular structure, so that the relationship between them can be reflected and mined through matrix operations.

[0127] The generated digital matrix is ​​subjected to XOR calculation. XOR operation (XOR) is a logical operation, and its rule is that when two input bits are different, the output is 1, and when they are the same, the output is 0. At each element position of the digital matrix, the corresponding digital characters (here the digital characters can be converted into binary form for XOR operation, for example, the number 5 is converted into binary 0101) are subjected to XOR operation in pairs according to the order of rows or columns. After such comprehensive XOR operation, the result matrix obtained is the XOR matrix. Due to the characteristics of XOR operation, the element values ​​in the XOR matrix finally obtained often show certain regularity, and only the first digit and the second digit (these two digits are 0 and 1 respectively) are included in this scheme. In one embodiment, the above XOR calculation can also be used to determine whether each character in the above matrix is ​​greater than a preset value, and if it is greater than, it is output as 1, and if it is not greater than, it is output as 0, thereby obtaining the above XOR matrix.

[0128] The XOR matrix with specific digital results obtained through the XOR operation further transforms and refines the original digital characters, providing a more distinctive and secure intermediate data form for subsequent combination with the character graph structure and generation of communication keys.

[0129] The above preset rules are to clarify how to reasonably and meaningfully combine the XOR matrix with the character graph structure. The above rules enable the digital information in the XOR matrix to interact with the character information in the character graph structure, thereby mining more relationships and features hidden in the data, providing richer materials for generating communication keys.

[0130] Assume that the preset rule is to superimpose in accordance with the corresponding node position. That is, for each node in the character graph structure, the element value of the corresponding position in the XOR matrix (which can be based on the row and column numbers of the node in the graph structure and the row and column numbers of the XOR matrix) is superimposed on the node. For example, if the character graph structure is a grid graph with 10 rows and 10 columns, and the XOR matrix is ​​also 10 rows and 10 columns, then the element value of the first row and first column of the XOR matrix is ​​superimposed on the node of the first row and first column of the character graph structure, and so on. Through such a superposition method, the digital information in the XOR matrix is ​​integrated into the character graph structure in a specific way, changing the attributes of some nodes in the character graph structure (which can be adding new numerical attributes, etc.), laying the foundation for the subsequent extraction of communication keys from this fused structure.

[0131] The main purpose of obtaining the two farthest first numbers in the XOR matrix and connecting them, as well as obtaining the two farthest second numbers and connecting them, is to find digital combinations with specific relationships in the XOR matrix, a structure obtained through special operations, and to make their relationships explicit by connecting them.

[0132] For an XOR matrix, to determine the two farthest first numbers, you can calculate the coordinate position of each first number in the matrix, and then use a distance formula (such as the Euclidean distance formula) to measure the distance between each first number to find the two farthest first numbers. For example, assuming the first number is 1, there is a 1 at a certain position (2,3) in the XOR matrix, and there is also a 1 at another position (8,9). By calculating the distance between them, it is found that these two are the farthest apart, so the positions of these two 1s are connected to obtain the first line. In the same way, for the second number, the two farthest second numbers are found through similar distance calculations and connected to obtain the second line.

[0133] The purpose of searching the edge intersecting the first line and / or the second line in the character graph structure as the target edge is to transfer the specific digital relationship embodied by the lines in the XOR matrix to the character graph structure, and to further mine the elements related to the generation of the communication key through the intersection relationship with the edges in the character graph structure.

[0134] After the first and second lines have been obtained in the XOR matrix, for the character graph structure, it is necessary to traverse all its edges to determine whether each edge intersects with the first and / or second lines. For example, check the node positions corresponding to the two endpoints of each edge one by one to see if there is a node position in the area passed by the first or second line. If there is such an edge, it is determined as the target edge. Through such a search process, the target edge related to the key line of the XOR matrix is ​​screened out from the character graph structure to prepare for the next step of generating the communication key.

[0135] Finally, the assignments on the target edge are combined to generate the communication key because the assignments on the target edge are determined based on the character relationships on adjacent nodes in the character graph structure, and the character graph structure integrates the comprehensive information set, the adjusted key parameters, and the relevant information in the fault prediction results. Therefore, the assignments on the target edge actually contain a variety of information closely related to the original data. By combining these assignments, a unique communication key that is deeply associated with the original data can be generated. This key can not only be used to encrypt and decrypt the original data, but also, due to its close connection with the original data, it guarantees the effectiveness and security of the encryption and decryption process to a certain extent, so that only the party who has the correct communication key can correctly process the relevant data.

[0136] In one embodiment, the method further comprises:

[0137] The comprehensive information set, feature importance sequence and optimized key parameters are compressed and format converted to obtain compact format data. The data compression adopts a hybrid compression algorithm based on dictionary learning and predictive coding. The data feature dictionary is first constructed through dictionary learning, and then the dictionary index is encoded and compressed using predictive coding. This greatly reduces data storage space while reducing information loss during data format conversion, thereby improving data storage and transmission efficiency.

[0138] The compact format data is encrypted based on a quantum encryption algorithm to obtain an encrypted data block. The quantum encryption algorithm uses the superposition and entanglement of quantum states to generate quantum keys to encrypt the compact format data. The transmission of quantum keys during the encryption process ensures the absolute security of key distribution based on the un-eavesdropping nature of the quantum channel. The encrypted data has extremely high anti-cracking capabilities, effectively ensuring the confidentiality of the data.

[0139] The encrypted data block is fragmented and redundantly encoded to obtain redundant fragment data. That is, the encrypted data block is divided into multiple data fragments, and each fragment is redundantly encoded, such as using the Reed-Solomon encoding algorithm to add redundant information, so that even if some fragments are lost or damaged, the original data can be restored through the redundant information, thereby improving the reliability and fault tolerance of data storage.

[0140] The redundant fragmented data is distributed and stored in multiple heterogeneous storage nodes to obtain storage configuration information, where multiple heterogeneous storage nodes include local storage devices, private cloud storage nodes with different architectures, and public cloud storage service nodes. Dynamic storage allocation strategies are formulated based on the performance, security, cost and other factors of each node to disperse the redundant fragmented data. The above method reduces the risk of data loss due to failure or attack of a single storage node. At the same time, the storage configuration information records the mapping relationship between data fragments and storage nodes to facilitate data reading and reorganization.

[0141] Reference Figure 2 In another embodiment of the present invention, a system for predicting faults based on electromechanical equipment operation information is provided, comprising:

[0142] The acquisition module is used to collect the operation information, operation condition information and environmental interference information of the electromechanical equipment and integrate the data to obtain a comprehensive information set;

[0143] An evaluation module, used for performing feature importance evaluation on the comprehensive information set based on information entropy to obtain a feature importance sequence;

[0144] An adjustment module, used for adaptively adjusting key parameters of a fault prediction algorithm based on a genetic algorithm based on the feature importance sequence to obtain an optimized fault prediction algorithm;

[0145] The prediction module is used to input the operation information of the electromechanical equipment into the optimized fault prediction algorithm to perform fault prediction and obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the fault location.

[0146] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0147] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0148] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0149] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0150] In summary, the method and system for fault prediction based on the operation information of electromechanical equipment provided in the embodiment of the present invention include: collecting the operation information, operation condition information and environmental interference information of the electromechanical equipment and integrating the data to obtain a comprehensive information set; performing feature importance evaluation based on information entropy on the comprehensive information set to obtain a feature importance sequence; based on the feature importance sequence, adaptively adjusting the key parameters of the fault prediction algorithm based on the genetic algorithm to obtain an optimized fault prediction algorithm; inputting the operation information of the electromechanical equipment into the optimized fault prediction algorithm for fault prediction to obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the location of the failure. In the present invention, based on the acquired operation information, operation condition information and environmental interference information of the electromechanical equipment, the key parameters of the fault prediction algorithm based on the genetic algorithm are adaptively adjusted so that it can adapt to the current application scenario. The defect that the current fault prediction algorithm lacks adaptability to the operation characteristics, working condition changes and environmental differences of different electromechanical equipment is overcome.

[0151] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0152] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0153] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for fault prediction based on electromechanical equipment operation information, characterized in that: The following steps are involved: Collect the operation information, operation condition information and environmental interference information of electromechanical equipment and integrate the data to obtain a comprehensive information set; Performing feature importance evaluation on the comprehensive information set based on information entropy to obtain a feature importance sequence; Based on the feature importance sequence, key parameters of the fault prediction algorithm based on the genetic algorithm are adaptively adjusted to obtain an optimized fault prediction algorithm; The operation information of the electromechanical equipment is input into the optimized fault prediction algorithm to perform fault prediction to obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the fault location.

2. The method for fault prediction based on electromechanical equipment operation information according to claim 1, characterized in that: The data integration includes: classifying and integrating the electromechanical equipment operation information collected from different sensors, the environmental interference information reflecting the environmental conditions of the equipment, and the operating condition information characterizing the current operating mode of the electromechanical equipment according to the data type and source to form a comprehensive information set.

3. The method for fault prediction based on electromechanical equipment operation information according to claim 1, characterized in that: The comprehensive information set is evaluated for feature importance based on information entropy to obtain a feature importance sequence, including: Calculate the information entropy of each information in the comprehensive information set; The degree of uncertainty of each piece of information is determined according to the size of the information entropy, and the feature importance sequence is obtained according to the degree of uncertainty.

4. The method for fault prediction based on electromechanical equipment operation information according to claim 1, characterized in that: The adaptive adjustment of key parameters of the fault prediction algorithm based on the genetic algorithm includes: The key parameters in the fault prediction algorithm are encoded as chromosomes in the genetic algorithm, and the feature importance sequence is used as the input of the fitness function in the genetic algorithm. The key parameters on the chromosome are continuously optimized through the selection, crossover and mutation operations of the genetic algorithm until the preset convergence conditions are met.

5. The method for fault prediction based on electromechanical equipment operation information according to claim 1, characterized in that: Inputting the operation information of the electromechanical equipment into the optimized fault prediction algorithm to perform fault prediction and obtain a fault prediction result, including: The operation information is divided into a plurality of information segments with time series characteristics, and a feature reconstruction process based on singular value decomposition is performed on each information segment to obtain a reconstructed feature matrix; Each reconstructed feature matrix is ​​input into the support vector machine classifier in the optimized fault prediction algorithm, and the reconstructed feature matrix is ​​mapped to a high-dimensional space using the kernel function of the support vector machine classifier. The optimal classification hyperplane is found in the high-dimensional space, and the fault classification result corresponding to each information fragment is calculated; Based on the dynamic time warping algorithm, the fault classification results of each information segment are integrated and corrected in time series to obtain the final fault prediction result; wherein, the dynamic time warping algorithm calculates the optimal matching path between the fault classification result sequences of different information segments, adjusts the classification result deviation caused by time difference, and thus determines the fault prediction result of the electromechanical equipment on the overall operation timeline.

6. The method for fault prediction based on electromechanical equipment operation information according to claim 1, characterized in that: After the fault prediction result is obtained, the following steps are included: Adding the comprehensive information set, the adjusted key parameters, and the characters in the fault prediction results to a preset graph structure template; the graph structure template includes a plurality of nodes, and adjacent nodes are connected by edges; According to the characters on the adjacent nodes, the values ​​on the edges connecting the adjacent nodes are determined to obtain the character graph structure; generating a communication key based on the character graph structure; The comprehensive information set, the adjusted key parameters, and the fault prediction results are encrypted based on the communication key and sent to the management terminal. Adding the comprehensive information set, the adjusted key parameters, and the characters in the fault prediction results to a preset graph structure template includes: Respectively obtaining the ratio of the number of characters in the comprehensive information set, the adjusted key parameters, and the fault prediction result; According to the ratio, the preset graph structure template is divided into regions to obtain a first region, a second region, and a third region; The characters in the comprehensive information set are added to each node in the first area, the characters in the adjusted key parameters are added to each node in the second area, and the characters in the fault prediction results are added to each node in the third area.

7. The method for fault prediction based on electromechanical equipment operation information according to claim 6, characterized in that: Generating a communication key based on the character graph structure includes: Respectively obtaining the comprehensive information set, the adjusted key parameters, and the numeric characters in the fault prediction results; Add the acquired digital characters to the matrix in sequence to generate a digital matrix; perform XOR calculation on the digital matrix to obtain an XOR matrix; the XOR matrix only includes the first number and the second number; According to a preset rule, superimposing the XOR matrix into the character graph structure; Obtain two farthest first numbers in the XOR matrix and connect them to obtain a first connection line; obtain two farthest second numbers in the XOR matrix and connect them to obtain a second connection line; Searching for an edge intersecting the first connecting line and / or the second connecting line in the character graph structure as a target edge; The assignments on the target edges are combined to obtain the communication key.

8. A system for fault prediction based on electromechanical equipment operation information, characterized in that: include: The acquisition module is used to collect the operation information, operation condition information and environmental interference information of the electromechanical equipment and integrate the data to obtain a comprehensive information set; An evaluation module, used for performing feature importance evaluation on the comprehensive information set based on information entropy to obtain a feature importance sequence; An adjustment module, used for adaptively adjusting key parameters of a fault prediction algorithm based on a genetic algorithm based on the feature importance sequence to obtain an optimized fault prediction algorithm; The prediction module is used to input the operation information of the electromechanical equipment into the optimized fault prediction algorithm to perform fault prediction and obtain a fault prediction result; wherein the fault prediction result includes the probability of failure of the electromechanical equipment, the fault type and the fault location.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images