A vacuum pump real-time data anomaly detection method and system

Abstract

The present invention provides a vacuum pump real-time data anomaly detection method and system, which relates to the field of intelligent control technology. The method includes: analyzing the physical phenomenon of the vacuum pump in the current working state according to the working state data to obtain the analysis result; according to the analysis result, for each data point p in the working state data, find the corresponding k points, and calculate the corresponding distance according to the k points; for any two data points p and data point o in the working state data, the reachable distance from the data point p to the data point o is defined as the maximum value between the k-distance of the data point o and the actual distance from the data point p to the data point o; for each data point p, calculate the inverse of the sum of the reachable distances of its k nearest neighbors to obtain the local reachable density; for each data point p, calculate the average value of the local reachable density of its k nearest neighbors. The present invention can improve the accuracy and efficiency of monitoring.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a method and system for detecting real-time data anomalies of a vacuum pump. Background Art

[0002] In the existing technology, although there are some methods that can monitor and analyze the working data of vacuum pumps in real time, these methods do have certain limitations in dealing with complex data processing and potential anomaly identification. These limitations become particularly obvious when dealing with large amounts of high-dimensional real-time data.

[0003] The amount of real-time data generated by vacuum pumps is relatively large, and the dimensions are diverse, including multiple parameters such as vacuum degree, temperature, pressure, vibration frequency, etc. However, the accuracy of existing methods in identifying abnormalities in vacuum pumps needs to be improved. Faced with high-dimensional real-time data, how to accurately distinguish normal fluctuations from abnormal states is a key issue. Traditional methods may rely too much on simple threshold judgments, which are often not flexible and accurate enough when faced with complex and changeable data. In addition, some potential anomalies may be hidden in seemingly normal data fluctuations, which requires more sophisticated anomaly detection technology for identification and early warning. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for detecting anomalies in real-time data of a vacuum pump, which can improve the accuracy and efficiency of monitoring to ensure that a rapid response can be made when any anomaly is detected.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a method for detecting anomalies in real-time data of a vacuum pump is provided, the method comprising:

[0007] Real-time collection of vacuum pump working status data, including vacuum degree, temperature, pressure, and vibration frequency;

[0008] Transmit working status data to the data processor in real time via data line or wireless network;

[0009] According to the working status data, the physical phenomena of the vacuum pump in the current working state are analyzed to obtain the analysis results. The physical phenomena include stress distribution, temperature change and pressure fluctuation;

[0010] According to the analysis results, for each data point p in the working status data, find the k nearest neighbors, sort these neighbors from near to far, calculate the k-distance, and record the distance from the data point p to its kth nearest neighbor as the k-distance of p; calculate the reachable distance from p to each neighbor o, for each neighbor o of the data point p, obtain the k-distance of the data point o, and calculate the actual distance from the data point p to the neighbor o; the reachable distance from the data point p to the data point o is defined as the maximum value between the k-distance of the data point o and the actual distance from the data point p to the data point o;

[0011] For each data point p, the local reachability density is obtained by calculating the inverse of the sum of the reachability distances of its k nearest neighbors; for each data point p, the average of the local reachability densities of its k nearest neighbors is calculated; the average is divided by the local reachability density of the data point p itself to obtain the local anomaly factor score of the data point p;

[0012] According to the local anomaly factor score, when an abnormal situation is detected, the early warning mechanism is triggered.

[0013] Furthermore, the working status data is transmitted to the data processor in real time via a data line or a wireless network, including:

[0014] Process the working status data and extract key features, including switch status, operating time, vacuum degree, pumping rate, ambient temperature, humidity, lubricating oil status, filter cleanliness, abnormal vibration and high temperature alarm;

[0015] According to key features, the working status data is divided into different categories to obtain classification results. The different categories include operation data, performance data, environmental data, maintenance data, and fault and warning data;

[0016] According to the classification results, each category of data is packaged into different data packets, and corresponding metadata is attached to each data packet, including timestamp and data source information;

[0017] For each data packet, the urgency of each data packet is evaluated according to the data category and specific content contained therein, so as to obtain an urgency evaluation result;

[0018] According to the evaluation result of urgency, the transmission priority of each data packet is set;

[0019] According to the priority of the data packet, the data packet is transmitted to the data processor in real time through the data line or wireless network.

[0020] Furthermore, the working status data is classified into different categories according to the key features to obtain classification results, including:

[0021] Encode the key features and randomly generate an initial population, where each individual in the initial population represents a combination of key features;

[0022] Define a fitness function to evaluate the contribution of each key feature combination to classification performance;

[0023] According to the fitness function, the corresponding individuals are determined to enter the next generation, and two individuals are randomly selected for crossover operation to generate a new combination of key features;

[0024] The gene positions of individuals are mutated with a certain probability, and the genetic operation is repeated until the preset number of iterations is reached to obtain the final combination of key features;

[0025] According to the preset classifier and the final combination of key features, the working status data is divided into different categories to obtain the classification results.

[0026] Furthermore, the key features are encoded and an initial population is randomly generated, including:

[0027] Determine a list of key characteristics, including switch status, operating hours, vacuum level, pumping rate, ambient temperature, humidity, lubricant condition, filter cleanliness, abnormal vibration and high temperature alarm;

[0028] According to the key feature list, determine the number of key features and the encoding length of each key feature;

[0029] According to the encoding length of each key feature, the selection state of each key feature is converted into a corresponding binary code;

[0030] Set the number of individuals in the population, and for each individual, randomly generate a binary string equal to the number of key features;

[0031] Combine all randomly generated individuals into an initial population.

[0032] Furthermore, for each data packet, the urgency of each data packet is evaluated according to the data category and specific content contained therein, so as to obtain an urgency evaluation result, including:

[0033] Analyze the nature and importance of each data item in each data packet and assign a weight to each data item;

[0034] Assign an urgency level to each data item based on its time sensitivity and immediacy factors;

[0035] For each data item, multiply its weight by the value of its urgency level to obtain the urgency score of the data item;

[0036] The urgency scores of all data items are added together to obtain the urgency evaluation result of the entire data packet.

[0037] Furthermore, the nature and importance of each data item in each data packet are analyzed, and a weight is assigned to each data item, including:

[0038] Classify all data items according to the physical quantities or states they reflect to obtain classification results;

[0039] For each data item, analyze the properties it represents and evaluate the importance of each data item to the vacuum pump operating status monitoring;

[0040] A weight is assigned to each data item according to its importance to the monitoring of the operating status of the vacuum pump.

[0041] Furthermore, according to the working state data, the physical phenomena of the vacuum pump in the current working state are analyzed to obtain analysis results. The physical phenomena include stress distribution, temperature change and pressure fluctuation, including:

[0042] Extract the vibration frequency data related to stress from the working status data; establish the mechanical model of the vacuum pump according to the structural design of the vacuum pump; calculate the stress distribution of the vacuum pump under the current working state according to the vibration frequency data and the mechanical model of the vacuum pump to obtain the stress distribution analysis results;

[0043] Extract temperature data from the working status data, and draw a temperature curve of temperature change over time based on the temperature data and the corresponding timestamp; detect abnormal points in the temperature curve through a preset temperature threshold; analyze the thermal performance of the vacuum pump during operation based on the temperature change curve and abnormal points to obtain the temperature change analysis results;

[0044] Extract pressure data from working status data; calculate the coefficient of variation of pressure data and draw a pressure fluctuation curve of pressure changing with time; detect abnormal fluctuations in the pressure curve through a preset fluctuation threshold; analyze pressure-related problems of the vacuum pump during operation according to the pressure fluctuation curve and abnormal fluctuation points to obtain pressure fluctuation analysis results;

[0045] The stress distribution analysis results, temperature change analysis results and pressure fluctuation analysis results are comprehensively analyzed to obtain a comprehensive analysis result, which includes existing failure modes, performance degradation trends or safety hazards.

[0046] In a second aspect, a vacuum pump real-time data anomaly detection system is used in the method described, comprising:

[0047] The acquisition module is used to collect the working status data of the vacuum pump in real time, including vacuum degree, temperature, pressure, and vibration frequency;

[0048] A sending module, used for transmitting the working status data to the data processor in real time via a data line or a wireless network;

[0049] An analysis module is used to analyze the physical phenomena of the vacuum pump in the current working state according to the working state data to obtain analysis results, the physical phenomena including stress distribution, temperature change and pressure fluctuation;

[0050] The first calculation module is used to find k nearest neighbors for each data point p in the working status data according to the analysis results, sort these neighbors from near to far according to the distance, calculate the k-distance, and record the distance from the data point p to its kth nearest neighbor as the k-distance of p; calculate the reachable distance from p to each neighbor o, for each neighbor o of the data point p, obtain the k-distance of the data point o, and calculate the actual distance from the data point p to the neighbor o; the reachable distance from the data point p to the data point o is defined as the maximum value between the k-distance of the data point o and the actual distance from the data point p to the data point o;

[0051] The second calculation module is used to calculate the inverse of the sum of the reachable distances of its k nearest neighbors for each data point p to obtain the local reachable density; for each data point p, calculate the average value of the local reachable densities of its k nearest neighbors; divide the average value by the local reachable density of the data point p itself to obtain the local anomaly factor score of the data point p;

[0052] The detection module is used to trigger the early warning mechanism when an abnormal situation is detected based on the local abnormal factor score.

[0053] According to a third aspect, a computing device includes:

[0054] one or more processors;

[0055] The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method described.

[0056] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method described is implemented.

[0057] The above solution of the present invention includes at least the following beneficial effects:

[0058] By collecting key working status data such as vacuum degree, temperature, pressure, vibration frequency, etc. in real time, the real-time operating status of the vacuum pump can be fully understood.

[0059] By analyzing the working status data, physical phenomena such as stress distribution, temperature changes and pressure fluctuations of the vacuum pump under the current working state can be revealed, which helps to timely discover potential performance problems and safety hazards.

[0060] By calculating the local reachable density and local anomaly factor scores of data points, this method can accurately identify abnormal points in the data, which may correspond to abnormal conditions or precursors of failure of the vacuum pump. When an abnormal condition is detected, the timely triggering of the early warning mechanism can help operators respond quickly, prevent the failure from expanding, and ensure the stable operation of the vacuum pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a flow chart of a method for detecting anomalies in real-time data of a vacuum pump provided in an embodiment of the present invention.

[0062] Figure 2 It is a schematic diagram of a vacuum pump real-time data anomaly detection system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in a form and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0064] like Figure 1 As shown, an embodiment of the present invention provides a method for detecting abnormalities in real-time data of a vacuum pump, the method comprising the following steps:

[0065] Step 11, collecting the working status data of the vacuum pump in real time, including vacuum degree, temperature, pressure, and vibration frequency;

[0066] Step 12, transmitting the working status data to the data processor in real time via a data line or a wireless network;

[0067] Step 13, analyzing the physical phenomena of the vacuum pump in the current working state according to the working state data to obtain analysis results, the physical phenomena including stress distribution, temperature change and pressure fluctuation;

[0068] Step 14: according to the analysis results, for each data point p in the working status data, find the corresponding k points, and calculate the corresponding distance based on the k points; for any two data points p and data point o in the working status data, the reachable distance from data point p to data point o is defined as the maximum value between the k-distance of data point o and the actual distance from data point p to data point o;

[0069] Step 15, for each data point p, the local reachable density is obtained by calculating the inverse of the sum of the reachable distances of its k nearest neighbors; for each data point p, the average value of the local reachable densities of its k nearest neighbors is calculated; the average value is divided by the local reachable density of the data point p itself to obtain the local anomaly factor score of the data point p;

[0070] Step 16: According to the local anomaly factor score, when an abnormal situation is detected, an early warning mechanism is triggered.

[0071] In an embodiment of the present invention, step 11, real-time collection of key working status data can ensure timely acquisition of the operating status of the vacuum pump. Step 12, data transmission through a data line or wireless network ensures the real-time and accuracy of the data, while improving the flexibility and scalability of the data acquisition system, so that the data can be quickly processed and analyzed. Step 13, through the analysis of physical phenomena, the working mechanism and performance status of the vacuum pump can be deeply understood, which is helpful to discover potential problems and hidden dangers. Step 15, by calculating the distance and reachable distance between data points, the similarity and difference between data points can be quantitatively evaluated. Step 14, by calculating the local reachable density and the local anomaly factor score, the abnormal points or outliers in the data set can be accurately identified. These points may correspond to abnormal conditions or fault signals of the vacuum pump, which helps to timely discover and deal with potential problems. Step 16, the timely triggering of the early warning mechanism can help operators respond quickly to abnormal conditions, prevent the expansion of faults or cause more serious consequences, and improve the operating safety and reliability of the vacuum pump.

[0072] In another preferred embodiment of the present invention, the above step 11 may include:

[0073] Real-time collection of vacuum pump working status data, including vacuum degree, temperature, pressure and vibration frequency. Among them, vacuum degree is a key indicator to measure the performance of vacuum pumps, which indicates the rarefaction of air in the pump. By real-time monitoring of vacuum degree, you can understand the working efficiency of the pump and whether there are problems such as leakage. Temperature is the heat generated by the vacuum pump during operation. Excessive temperature may cause equipment damage or performance degradation. Therefore, real-time monitoring of temperature data is crucial to prevent overheating and protect equipment. Pressure data reflects the changes in gas pressure during the operation of the vacuum pump. Abnormal pressure fluctuations may indicate pump failure or performance problems. The vibration frequency of the vacuum pump can reflect the working status of its internal mechanical components. Abnormal vibration may mean bearing wear, imbalance or other mechanical failures. By monitoring the vibration frequency, these problems can be discovered and dealt with in a timely manner.

[0074] For example, a manufacturing company uses a vacuum pump equipped with sensors and a monitoring system. These sensors can collect real-time data on the vacuum pump's vacuum level, temperature, pressure, and vibration frequency.

[0075] During a production process, the monitoring system suddenly discovered that the vibration frequency of a vacuum pump had abnormal fluctuations. Further analysis revealed that this abnormal fluctuation was consistent with signs of bearing wear. As a result, the company immediately arranged for maintenance and overhaul of the vacuum pump to avoid possible serious failures.

[0076] By discovering and handling the problem in time, the company avoided production interruptions and product quality problems caused by vacuum pump failure, and also extended the service life of the equipment. This case fully demonstrates the importance of real-time collection of working status data in preventing equipment failure and improving production efficiency.

[0077] In a preferred embodiment of the present invention, the above step 12, transmitting the working status data to the data processor in real time via a data line or a wireless network, may include:

[0078] Step 121, processing the working status data to extract key features, the key features including switch status, operating time, vacuum degree, pumping rate, ambient temperature, humidity, lubricating oil status, filter cleanliness, abnormal vibration and high temperature alarm;

[0079] Step 122, classifying the working status data into different categories according to the key features to obtain classification results, the different categories including operation data, performance data, environment data, maintenance data, and fault and warning data;

[0080] Step 123, according to the classification result, the data of each category are packaged into different data packets, and corresponding metadata including timestamp and data source information are attached to each data packet;

[0081] Step 124, for each data packet, the urgency of each data packet is evaluated according to the data category and specific content contained therein, so as to obtain an evaluation result of the urgency;

[0082] Step 125, setting a transmission priority for each data packet according to the evaluation result of the urgency;

[0083] Step 126 , transmitting the data packet to the data processor in real time via a data line or a wireless network according to the priority of the data packet.

[0084] In an embodiment of the present invention, step 121, by extracting key features, such as switch status, operating time, etc., the redundancy and complexity of data can be reduced, making subsequent data processing and analysis more efficient. At the same time, these key features are also important indicators for understanding the working status and performance of the vacuum pump, which helps to detect abnormal situations in a timely manner. Step 122, classifying the data helps to better organize and understand the data, so that different types of data can be processed and analyzed in a targeted manner. For example, operating data can be used to analyze operating habits, performance data can be used to evaluate the working efficiency of the vacuum pump, and environmental data can help understand the impact of environmental factors on equipment performance. Step 123, the creation of data packets and metadata improves the traceability and management efficiency of data. Metadata such as timestamps and data source information help determine the accuracy and reliability of data, as well as the source and generation background of the data. Step 124, the assessment of the degree of urgency helps to determine the priority of data processing and ensure that important or urgent data can be processed in a timely manner. For example, fault and warning data may require a faster response speed to deal with potential problems in a timely manner. Step 125, setting transmission priorities ensures the timely transmission and processing of key data. In the case of limited network resources or large data transmission volume, this priority setting can ensure that important data will not be delayed, thereby improving the response speed and efficiency of the entire system. Step 126, data transmission according to priority not only ensures timely processing of key data, but also optimizes the utilization of network resources. Real-time transmission ensures the timeliness and accuracy of data.

[0085] The above step 121 may include:

[0086] Obtain raw operating status data from the vacuum pump's control system or sensor network; read the vacuum pump's switch status (on or off) from the control signal or status indicator; record the vacuum pump's operating time from startup to the current moment, which can be calculated using a timestamp; obtain real-time vacuum readings from the vacuum sensor; calculate or directly read the pumping rate from the relevant sensor based on the vacuum pump's operating principle and sensor data; obtain current ambient temperature and humidity data through environmental sensors; evaluate the lubricating oil status (such as cleanliness, viscosity, etc.) through lubricating oil sensors or regular oil quality analysis; determine the filter's blockage or cleanliness through a differential pressure sensor or regular physical inspection; use a vibration sensor to monitor the vacuum pump's vibration; read data from a temperature sensor and set a threshold to trigger a high temperature alarm; perform unit conversion, standardization, or normalization on the extracted feature data; apply filtering algorithms (such as moving average) to smooth the data and reduce noise interference; store the extracted key feature data in an appropriate database or data structure.

[0087] In a preferred embodiment of the present invention, the above step 122 classifies the working status data into different categories according to the key features to obtain classification results, including:

[0088] Step 1221, encoding the key features and randomly generating an initial population, each individual in the initial population represents a combination of key features;

[0089] Step 1222, defining a fitness function for evaluating the contribution of each key feature combination to classification performance;

[0090] Step 1223, according to the fitness function, determine the corresponding individuals to enter the next generation, randomly select two individuals to perform a crossover operation, and generate a new combination of key features;

[0091] Step 1224, mutate the gene positions of the individual with a certain probability, and repeat the genetic operation until a preset number of iterations is reached to obtain the final combination of key features;

[0092] Step 1225, classify the working status data into different categories according to the preset classifier and the final combination of key features to obtain a classification result.

[0093] In an embodiment of the present invention, step 1221, by encoding the key features, they can be converted into a form suitable for genetic algorithm processing. The randomly generated initial population increases the diversity of the search space, which helps the algorithm to find the optimal solution in the global scope, rather than falling into the local optimum. Step 1222, the fitness function is the key to the optimization process in the genetic algorithm, which quantifies the contribution of each key feature combination to the classification performance. Through this function, the algorithm can select and optimize the feature combinations that have the greatest impact on the classification results in a targeted manner. Step 1223, the crossover operation is one of the core steps in the genetic algorithm. It generates new individuals that may have better performance by combining the excellent genes (i.e., key features) of different individuals. This helps the algorithm to quickly find a better solution in the search space. Step 1224, the mutation operation increases the diversity of the population, which helps the algorithm to jump out of the local optimal solution and explore a wider search space. At the same time, it also increases the possibility of finding the global optimal solution. Step 1225, through the optimization of the previous steps, the key feature combination that contributes the most to the classification performance is obtained. Using these features for classification can improve the accuracy and efficiency of classification.

[0094] In a preferred embodiment of the present invention, the above step 1221, encoding the key features and randomly generating an initial population, includes:

[0095] Step 12211, determine the key feature list, including switch status, operating time, vacuum degree, pumping rate, ambient temperature, humidity, lubricating oil status, filter cleanliness, abnormal vibration and high temperature alarm;

[0096] Step 12212, determining the number of key features and the encoding length of each key feature according to the key feature list;

[0097] Step 12213, converting the selection state of each key feature into a corresponding binary code according to the code length of each key feature;

[0098] Step 12214, setting the number of individuals in the population, and for each individual, randomly generating a binary string equal to the number of key features, specifically including: randomly generating the initial population ensures the diversity of the population, which helps the genetic algorithm find the global optimal solution during the search process, and setting a suitable population size can ensure the search efficiency while avoiding the waste of computing resources;

[0099] Step 12215, combining all randomly generated individuals into an initial population, specifically including: the formation of the initial population provides a starting point for subsequent genetic algorithm operations. Through the optimization process of the genetic algorithm, a feature combination with better performance can be evolved from the initial population, thereby improving the accuracy of the classification task.

[0100] In the embodiment of the present invention, step 12211 helps to reduce the dimension of the data. By eliminating irrelevant or redundant features, the impact of noise on classification performance can be reduced. Step 12212, the appropriate encoding length can reduce the computational complexity while ensuring the integrity of feature information. Step 12213, for each key feature, it is converted into a corresponding binary code according to whether it is selected (or the range of values). The binary code simplifies the crossover and mutation operations in the genetic algorithm, and the encoded features are easier to process and optimize in the genetic algorithm.

[0101] In a preferred embodiment of the present invention, in the above step 1222, the calculation formula of the fitness function is:

[0102] ;

[0103] in, Represents the true positive, that is, the number of samples that are actually positive and correctly classified as positive. Represents true negative examples, that is, the number of samples that are actually negative and correctly classified as negative. Represents false positives, that is, the number of samples that are actually negative but are mistakenly classified as positive. Represents false negative examples, that is, the number of samples that are actually positive but are mistakenly classified as negative. , , and is the weight parameter. The purpose of this fitness function is to gradually optimize the combination of key features through the selection, crossover, and mutation operations of the genetic algorithm, thereby improving the performance of the classifier. By continuously evaluating the fitness of each generation of individuals, the optimal feature combination can be gradually found, thereby improving the classification accuracy and reliability of the vacuum pump working status data.

[0104] When specifically applied, this includes:

[0105] Step 1221, encoding the key features and randomly generating an initial population;

[0106] Step 1222, according to the specific requirements of the problem, design a fitness function that can evaluate the contribution of key feature combinations to classification performance.

[0107] Step 1223, using the roulette wheel selection method, select individuals with better performance from the current population according to the value of the fitness function; the selection process should ensure that excellent individuals have a higher probability of being selected to pass on their excellent genes to the next generation. Randomly pair the selected individuals to form parent pairs. For each pair of parents, randomly select one or more crossover points, exchange genes at these points, and generate new offspring individuals. The crossover operation helps to combine the excellent genes of different parents, thereby generating new individuals that may have better performance.

[0108] Step 1224, for each newly generated offspring individual, its gene is randomly changed with a certain probability (such as flipping a bit in the binary code); the mutation operation increases the diversity of the population, helps the algorithm to jump out of the local optimal solution, repeats the selection, crossover and mutation operations to form a new generation of population, and records the individual with the highest fitness in each generation in order to track the optimization process of the algorithm. When the preset number of iterations is reached, the iteration is stopped.

[0109] Step 1225, obtain the optimized key feature combination from the output of the genetic algorithm. This combination is obtained through multiple rounds of genetic operations (selection, crossover, mutation), and is considered to be the feature set that contributes most to the classification performance.

[0110] In this step, the support vector machine (SVM) is selected as the classifier. After the classifier is determined, the SVM model is trained using the data set corresponding to the optimized key feature combination. This data set has been labeled so that the model can learn the mapping relationship from features to categories. During the training process, the parameters of the SVM (such as the penalty coefficient C, the kernel function type and its parameters, etc.) need to be adjusted to achieve the best classification performance.

[0111] After the classifier is trained, the original working status data is preprocessed and converted so that it can be input into the classifier for classification. The preprocessing step may include data cleaning (removing outliers, missing value processing, etc.), data standardization or normalization (making different features have the same scale), and selecting corresponding feature columns based on the optimized key feature combination. The conversion step involves converting the data from the original format to a format that the classifier can accept, such as converting text data to numerical data.

[0112] The preprocessed and converted working status data is input into the trained SVM model for classification prediction. The model will output the category label of each data point.

[0113] According to the output labels of the SVM model, the working status data can be divided into different categories, such as operation data, performance data, environmental data, maintenance data, and fault and warning data. For example, operation data may reflect the use of equipment, performance data may reveal the operating efficiency and stability of equipment, and fault and warning data may be used to timely discover and prevent potential problems.

[0114] In summary, the implementation process of step 1225 involves multiple links, such as using the feature combination optimized by the genetic algorithm, selecting a suitable classifier, training the classifier, preprocessing and converting data, applying the classifier for classification, and classifying the data according to the classification results. These links are interconnected and together constitute a complete processing flow from the original working state data to the classification results.

[0115] In another preferred embodiment of the present invention, the above step 123 may include:

[0116] The classification results of the working status data are obtained from step 1225. These data have been accurately divided into different categories, such as operation data, performance data, environmental data, maintenance data, and fault and warning data.

[0117] For each data category, create an empty data packet to store the data of that category. At the same time, define a metadata structure to record additional information for each data packet. The metadata structure should contain fields such as timestamp (recording the creation time of the data packet or the collection time of the data) and data source information (describing the source of the data, such as device ID, sensor number, etc.).

[0118] For each data point in the classification result, add it to the corresponding category of the data package according to its category. When adding data points, ensure the integrity of the data and the correct format. If the data needs to be further transformed or formatted to meet specific storage or transmission requirements, it is also completed in this step.

[0119] When a data packet of a certain category is filled, metadata is generated for the data packet. This includes assigning a value to the timestamp field in the metadata structure (usually using the current time or the earliest or latest time in the data point as the timestamp), and filling in the data source information (filled in according to the actual situation of the data, such as which device or sensor the data is collected from, etc.).

[0120] In a preferred embodiment of the present invention, the above step 124 evaluates the urgency of each data packet according to the data category and specific content contained therein to obtain an urgency evaluation result, including:

[0121] Step 1241, analyzing the nature and importance of each data item in each data packet, and assigning a weight to each data item;

[0122] Step 1242, assigning an urgency level to each data item based on the time sensitivity and immediacy factors of the data item, specifically including: dividing the urgency of the data item into several levels according to actual needs, such as "urgent", "high", "medium", "low", etc., each level should correspond to a specific numerical range; for each data item, assessing its time sensitivity. Time sensitivity refers to the rate at which the value of a data item changes over time. For example, some data may be extremely valuable when it is first generated, but its value decreases rapidly over time; while other data may maintain a stable value over a period of time; set a time window for the data item, and the value of the data item within this time window is considered to be the highest; assign a time sensitivity score to the data item based on the change in its value within the time window, and the higher the score, the more time sensitive the data item is; the immediacy factor reflects the degree to which the data item needs to be processed or paid attention to immediately, which is related to the real-time nature of the data, the update frequency, and the urgency of processing the data; understand the update frequency of the data item, such as real-time update, daily update, weekly update, etc.; consider the processing urgency and evaluate whether the data item needs to be processed immediately to avoid potential problems or seize important opportunities; assign an immediacy score to the data item based on its update frequency and processing urgency, and the higher the score, the stronger the immediacy demand for the data item; combine the time sensitivity score and the immediacy score of the data item, and use weighted average for comprehensive evaluation, and merge the two scores into one comprehensive score; set the comprehensive score threshold corresponding to different emergency levels according to the emergency level standards defined previously; compare the comprehensive score of each data item with the threshold to determine the emergency level to which it belongs;

[0123] Step 1243, for each data item, multiply its weight by the value of its urgency level to obtain the urgency score of the data item, specifically including: obtaining the weight and urgency level value of each data item, the weight is assigned according to the nature and importance of the data item in step 1241, and the urgency level value is determined according to time sensitivity and immediacy factors in step 1242; for each data item in the data packet, performing the following operations:

[0124] a. Get the weight value of the data item, denoted as weight.

[0125] b. Get the urgency level value of the data item, recorded as urgency_level, which is the specific number set in step 1242 according to the urgency level standard, such as "urgent" may be 4, "high" is 3, "medium" is 2, "low" is 1, or a more sophisticated numerical system is used.

[0126] c. Multiply the weight value by the urgency level value, and the calculation formula is: urgency_score = weight × urgency_level. The calculation result is the urgency score of the data item, recorded as urgency_score.

[0127] Associate the calculated urgency score urgency_score with the data item and record it. This can be achieved by adding a new field or attribute to each data item in the data structure; repeat the above steps until all data items in the data packet have been processed and each data item has a corresponding urgency score.

[0128] Step 1244, adding up the urgency scores of all data items to obtain the urgency assessment result of the entire data packet, specifically including: traversing each data item in the data packet, for each data item, obtaining its urgency score calculated in step 1243, and adding it to the urgency assessment result variable. When the urgency scores of all data items are added up, the values ​​in the urgency_score variables of all data items represent the urgency assessment result of the entire data packet.

[0129] In the embodiment of the present invention, data packets can be sorted according to their urgency evaluation results, and data packets with high urgency are processed first, thereby ensuring that key data is processed in a timely manner. By understanding the urgency of each data packet, processing resources can be allocated more reasonably to improve overall processing efficiency. The urgency evaluation results of data packets can provide support information for decision-making. By processing data packets with high urgency in a timely manner, the risks caused by data delays or omissions can be reduced.

[0130] In a preferred embodiment of the present invention, the above step 1241, analyzing the nature and importance of each data item in each data packet and assigning a weight to each data item, includes:

[0131] Step 12411, classify according to the physical quantity or state reflected by all data items to obtain the classification result, specifically including: obtaining a data packet, and each data item in the data packet is accessible, and the data items are stored in a specific format, such as CSV, JSON or XML; traversing each data item in the data packet, for each data item, identifying the physical quantity or state it represents, which can be achieved in the following ways: View metadata, data items contain metadata, such as tags or descriptions, which clearly indicate the physical quantity or state represented by the data item. Use naming conventions, if the naming of data items follows a specific convention, such as using "Temp" to represent temperature and "Press" to represent pressure, then the physical quantity or state can be identified by parsing the name of the data item.

[0132] After identifying the physical quantity or state represented by the data item, define the classification criteria. These criteria should be based on the nature of the physical quantity or state to ensure that data items of similar nature are classified into the same category. For example, all temperature-related data items can be classified into the "temperature" category, all pressure-related data items can be classified into the "pressure" category, and so on. Assigning each data item to a category based on the defined classification criteria can be achieved by creating a dictionary where the key is the category name and the value is a list of data items belonging to that category.

[0133] Step 12412, for each data item, analyze the properties it represents, and evaluate the importance of each data item to the monitoring of the operating status of the vacuum pump, specifically including: accessing each data item in the data package, ensuring that the information of the data item is readable and parsable, and extracting the metadata, name, value and any relevant descriptive information of the data item; analyzing the category to which the data item belongs (such as temperature, pressure, flow, etc.), which helps to understand the specific role of the data item in the monitoring of the operating status of the vacuum pump; considering the relationship between the data item and other data items, such as whether they are interdependent or independent; determining the nature of the physical quantity or state reflected by the data item, such as whether it is a critical operating parameter, whether it is safety-related, whether it directly affects the performance of the vacuum pump, etc.; analyzing the range of variation, normal values ​​and abnormal values ​​of the data item, and the possible impact of these values ​​on the operation of the vacuum pump; based on the analysis results of the properties of the data item, evaluating the importance of each data item to the monitoring of the operating status of the vacuum pump; using historical data to determine the importance level of the data item. For example, some key temperature or pressure readings may be identified as being essential for monitoring the status of a vacuum pump, while other auxiliary data may be considered less important. For example, for each data item Di, define the importance-related attributes of the data item:

[0134] Criticality , indicating whether the data item is a key parameter of the vacuum pump (1 is key, 0 is non-key).

[0135] Safety relevance , indicating whether the data item is directly related to security (1 for relevant and 0 for irrelevant).

[0136] Independence , indicating whether the data item is independent of other parameters (1 for independence and 0 for dependence).

[0137] Historical abnormality rate , represents the historical anomaly rate of the data item, ranging from 0 to 1, reflecting the frequency of anomalies in the data item in the past.

[0138] Based on the above attributes, data items The importance of monitoring the operating status of vacuum pumps It can be calculated by the following formula:

[0139] ;

[0140] in, , , , and is a weight parameter; if the data item is a key parameter, then =1, otherwise =0.

[0141] If the data item If it is related to safety, =1, otherwise =0.

[0142] Historical abnormality rate It indicates the historical anomaly rate of the data item, ranging from 0 to 1, reflecting the stability of the item; a higher anomaly rate may indicate that the fluctuation of the data item has a greater impact on monitoring.

[0143] If the data item Independent of other data items, =1, otherwise =0.

[0144] If the data item has a significant relationship with other key data items, then It will be assigned a certain value, reflecting its dependency on other key data items.

[0145] Step 12413, assign a weight to each data item according to its importance to the operating status monitoring of the vacuum pump, specifically including: determining weight assignment criteria, which should be based on the importance assessment results of the data items, and the weights can be continuous values ​​(such as real numbers between 0 and 1) or discrete levels (such as high, medium, and low); according to the importance of each data item assessed in step 12412, use the defined weight criteria to assign a specific weight value to each data item, ensuring that the weight assignment process is consistent and repeatable to avoid deviations caused by subjectivity.

[0146] In the embodiment of the present invention, by classification and weight allocation, data items that are critical to the operation status monitoring of the vacuum pump can be more accurately identified and processed, thereby improving the accuracy of data processing. By assigning different weights to different data items, processing resources, such as computing resources and human resources, can be more reasonably allocated to ensure priority processing of important data items. Through the accurate processing and analysis of the vacuum pump operation status monitoring data, potential problems can be discovered in time and corresponding measures can be taken, thereby improving the overall performance and stability of the vacuum pump system.

[0147] In a preferred embodiment of the present invention, the above step 13 analyzes the physical phenomena of the vacuum pump in the current working state according to the working state data to obtain analysis results, and the physical phenomena include stress distribution, temperature change and pressure fluctuation, including:

[0148] Step 131, extracting vibration frequency data related to stress from the working state data; establishing a mechanical model of the vacuum pump according to the structural design of the vacuum pump; calculating the stress distribution of the vacuum pump in the current working state according to the vibration frequency data and the mechanical model of the vacuum pump to obtain a stress distribution analysis result;

[0149] Step 132, extracting temperature data from the working status data, and drawing a temperature curve of temperature change over time according to the temperature data and the corresponding timestamp; detecting abnormal points in the temperature curve according to a preset temperature threshold; analyzing the thermal performance of the vacuum pump during operation according to the temperature change curve and the abnormal points to obtain a temperature change analysis result;

[0150] Step 133, extracting pressure data from the working status data; calculating the coefficient of variation of the pressure data, and drawing a pressure fluctuation curve of the pressure changing with time; detecting abnormal fluctuations in the pressure curve through a preset fluctuation threshold; analyzing pressure-related problems of the vacuum pump during operation according to the pressure fluctuation curve and abnormal fluctuation points, so as to obtain a pressure fluctuation analysis result;

[0151] Step 134 , comprehensively analyzing the stress distribution analysis results, the temperature change analysis results, and the pressure fluctuation analysis results to obtain a comprehensive analysis result, which includes the existing failure modes, performance degradation trends, or safety hazards.

[0152] In the embodiment of the present invention, by analyzing the working status data of the vacuum pump, the working status of the vacuum pump can be monitored in real time, abnormal conditions can be discovered in time, and possible faults can be warned in advance, so that fault warnings can be performed more accurately and reliably. By analyzing the working status data of the vacuum pump, the performance status of the equipment can be understood, the performance decline trend of the equipment can be discovered in time, and corresponding maintenance measures can be taken, so that the equipment maintenance can be performed more effectively and the maintenance efficiency and effect of the equipment can be improved. By analyzing the working status data of the vacuum pump, the operating status of the equipment can be monitored in real time, abnormal conditions of the equipment can be discovered in time, and corresponding treatment measures can be taken, so that the stable operation and safe operation of the equipment can be more effectively guaranteed.

[0153] In another preferred embodiment of the present invention, the above step 131 may include:

[0154] Accurately identify and extract stress-related vibration frequency data from the working status data. These data are collected by the vibration sensor installed on the vacuum pump and reflect the vibration of the vacuum pump during operation. When extracting data, pre-process the original signal, such as filtering and noise reduction, to ensure that the extracted vibration frequency data is accurate and effective.

[0155] According to the structural design of the vacuum pump, obtain or create the detailed geometry of its main components. This involves using CAD (computer-aided design) software to draw or import the various components of the vacuum pump. Make sure that the geometric model accurately reflects the actual structure of the vacuum pump, including all important details and features.

[0156] Assign the correct material properties to each component in the mechanical model of the vacuum pump. These properties include the elastic modulus (Young's modulus), Poisson's ratio, density, and any relevant material strength data (such as yield strength, tensile strength, etc.). Material properties can be obtained from the material database or determined through experimental testing. Discretizing the geometric model into a finite number of elements (usually quadrilateral or hexahedral elements) is a process called meshing or discretization. The density and type of mesh will affect the accuracy and computational efficiency of the analysis. In areas of stress concentration or complex geometry, a denser mesh is required.

[0157] Define the boundary conditions of the model, including constraints (such as fixed support points) and external loads (such as gravity, pressure, thermal loads, etc.). For a vacuum pump, important loads include internal pressure, thermal stress, and dynamic loads due to vibration.

[0158] According to the type of analysis (static analysis, dynamic analysis, thermal analysis, etc.), select a solver (for example, for static structural analysis, select a structural solver; for dynamic analysis, select the corresponding dynamic solver; for thermal analysis, select a thermal solver, etc.), set the parameters of the solver, and run the analysis using finite element analysis software (such as ANSYS). The software will solve the mechanical model according to the input conditions and output the stress, strain, displacement and other results of each element. Use post-processing tools to visualize and analyze the results, which include stress distribution diagrams, deformation diagrams, numerical data in reports, etc. Compare the analysis results with experimental data or known design criteria to verify the accuracy of the model. If necessary, optimize the design of the vacuum pump based on the analysis results, and then repeat the above finite element analysis process.

[0159] In another preferred embodiment of the present invention, the above step 132 may include:

[0160] Open the data file or database containing the working status data of the vacuum pump; traverse the data and locate the field or column where the temperature data is located; extract all temperature data points and the corresponding timestamps; ensure that the correspondence between the temperature data and the timestamp data is accurate; use a chart drawing library (such as Matplotlib, Plotly, etc.) to initialize a chart object, set the X-axis of the chart to the timestamp data and the Y-axis to the temperature data, select a chart type, such as a line chart, to intuitively display the trend of temperature changes over time, configure the chart title, axis labels, legends, etc. to enhance the readability of the chart; pass the temperature data and timestamp data to the chart object to draw a curve of temperature changes over time.

[0161] Set one or more temperature thresholds based on the normal operating temperature range or safety limits of the vacuum pump; traverse the temperature data and compare each temperature data point with the preset threshold. If a temperature data point exceeds the preset threshold range, mark it as an abnormal point and record the timestamps and temperature values ​​of all abnormal points for subsequent analysis.

[0162] Open the temperature-time curve that has been drawn, check the overall trend of the curve, pay attention to the peaks and valleys of the temperature, and observe whether the temperature shows periodic changes or whether there is a clear upward or downward trend. Mark the temperature peaks and valleys on the curve, record their timestamps and temperature values, and analyze whether these peaks and valleys appear at a certain time interval to identify periodic fluctuations; if the temperature shows a continuous upward or downward trend, try to find the time point when this change began and analyze the possible reasons.

[0163] Analyze the time, frequency and amplitude of abnormal points. According to the abnormal points marked before, count the total number of abnormal points and calculate the frequency of abnormal occurrence. Analyze the time period when the abnormal points appear to see if it is related to a specific operation stage or external environmental conditions. Compare the amplitude difference between the abnormal point temperature and the normal temperature fluctuation to quantify the severity of the abnormal point.

[0164] According to the records of the vacuum pump during operation, such as power on and off time, load changes, power output, etc., these operation records are compared with the time when the abnormal points occur to find possible causal relationships, obtain the internal sensor data of the vacuum pump (such as pressure, flow, etc.), and analyze the relationship between these data and temperature abnormal points.

[0165] Understand the working principle of the vacuum pump, including its internal heat generation and heat dissipation mechanism, analyze the impact of operating conditions such as ambient temperature, humidity, workload, etc. on the temperature of the vacuum pump, and infer the possible causes of normal and abnormal temperature fluctuations based on the working principle and operating conditions.

[0166] Analyze the potential damage of abnormally high temperature to the internal materials of the vacuum pump (such as seals, bearings, etc.), analyze the impact of abnormal temperature on the efficiency and working stability of the vacuum pump, and evaluate the degree of reduction in the service life of the vacuum pump due to long-term abnormal temperature based on historical data and literature.

[0167] For example, a vacuum pump works for 8 hours a day, starting at 8 am and shutting down at 4 pm. The abnormal points mainly occur between 2 pm and 4 pm. The load gradually increases during the working hours of the day and reaches a peak in the afternoon. The abnormal points coincide with the peak load time. The power output increases with the increase of load, and the power output when the abnormal points appear is the highest value of the day.

[0168] Internal sensor data is associated with anomalies:

[0169] When the pressure inside the vacuum pump increases, the temperature also increases. The pressure reading at the time of the abnormal point is higher than the average. The flow rate shows a downward trend before the abnormal point, which may indicate a blockage or leak inside the pump.

[0170] Working principle and operating conditions analysis:

[0171] When the vacuum pump is working, internal friction and compression process will generate heat. Usually, the heat dissipation system can effectively remove this heat. However, under high load and high temperature environment, the heat dissipation system may not work effectively. The operating environment temperature is 25°C and the humidity is 60%. These conditions are within the design operating range of the vacuum pump, so the external environment is unlikely to be the cause of the abnormality.

[0172] Potential damage from abnormally high temperatures:

[0173] Long-term high temperature may cause seal hardening and insufficient bearing lubrication, thus shortening its service life. The impact of abnormal temperature on vacuum pump efficiency and working stability: abnormally high temperature may cause the efficiency of the vacuum pump to decrease because it needs to consume more energy to maintain its operation. Impaired working stability: frequent temperature fluctuations may cause the operation of the vacuum pump to become unstable, increasing the risk of failure. According to historical data, the service life of vacuum pumps that work at abnormally high temperatures for a long time may be shortened by 30%-50%.

[0174] In another preferred embodiment of the present invention, the above step 133 may include:

[0175] Open the file or database containing the vacuum pump operating status data and locate the field or column storing the pressure data. Extract all pressure data points, making sure that each data point contains the corresponding timestamp.

[0176] Calculate the average value of the pressure data, calculate the absolute value of the difference between each pressure data point and the average value, calculate the average of these absolute differences, and obtain the average absolute deviation; divide the average absolute deviation by the average value of the pressure data to obtain the coefficient of variation, which reflects the degree of dispersion of the pressure data.

[0177] Use a chart drawing tool (such as Matplotlib, Seaborn, etc.) to create a time-pressure chart, where the X-axis represents the timestamp and the Y-axis represents the pressure value; select a chart type, such as a line chart, to clearly show the fluctuation of pressure over time, pass the extracted pressure data and timestamp to the chart tool, and draw a pressure fluctuation curve.

[0178] Set one or more pressure fluctuation thresholds, which can be based on the normal working pressure range of the vacuum pump or fluctuations in historical data. Traverse the pressure data and calculate the pressure change between adjacent data points. If a certain change exceeds the preset fluctuation threshold, mark the point as an abnormal fluctuation point. Observe the overall trend of the pressure fluctuation curve to identify periodic fluctuations or abnormal patterns. Analyze the time, frequency and amplitude of the abnormal fluctuation points, and their association with the operating status of the vacuum pump (such as power on and off, load changes, etc.). Combined with the working principle and operating conditions of the vacuum pump, explain the possible causes of abnormal fluctuations, such as seal leakage, air inlet blockage, internal wear of the pump, etc.

[0179] In another preferred embodiment of the present invention, the above step 134 may include:

[0180] Summarize the stress distribution analysis results, temperature change analysis results, and pressure fluctuation analysis results obtained in the previous steps; compare these results to find out the commonalities and differences between them, and pay special attention to those areas and time points that show abnormalities or deviate from the normal range in different analyses. Combined with the working principle and actual operation of the vacuum pump, analyze the possible causal relationship between abnormal stress distribution, abnormal temperature change, and abnormal pressure fluctuation. Identify the patterns that may cause vacuum pump failure, such as material fatigue caused by stress concentration, seal failure caused by excessive temperature, or unstable pump performance caused by excessive pressure fluctuation. According to the long-term trend of temperature change and pressure fluctuation, analyze whether the vacuum pump performance has a downward trend. Combined with the stress distribution, evaluate whether this downward trend may be caused by the gradual wear or deformation of the internal structure of the pump. Predict the consequences of performance degradation if no corresponding measures are taken, such as reduced efficiency, increased energy consumption, or increased maintenance costs. Analyze whether abnormal stress distribution, abnormal temperature increase, and abnormal pressure fluctuation may constitute safety hazards. Evaluate the safety accidents that these hazards may cause in extreme cases, such as pump body rupture, leakage caused by seal failure, or production interruption caused by sudden pump shutdown. Assign risk ratings to identified safety hazards based on historical data and failure cases of similar equipment.

[0181] In another preferred embodiment of the present invention, the above step 14 may include:

[0182] Choose an appropriate value of k. The value of k is a small positive integer that represents the number of nearest neighbors to consider. Use a data structure such as a KD tree, a ball tree, or a simple distance matrix to store the points in the work status data for efficient nearest neighbor search. For each data point p in the work status dataset, do the following.

[0183] Find the k nearest neighbors:

[0184] Using the chosen distance metric (e.g., Euclidean distance) and data structure, find the k nearest neighbors of data point p. Sort these neighbors by distance from closest to farthest.

[0185] Calculate k-distance:

[0186] The distance from data point p to its kth nearest neighbor is recorded as p's k-distance, which means that among all the neighbors of p, there are exactly k-1 points closer than this distance, and the rest are farther than this distance. For each data point p in the working state dataset, perform the operation again to calculate the reachable distance.

[0187] Calculate the reachable distance from p to each neighbor o:

[0188] For each neighbor o of data point p (you can start with the k nearest neighbors found in step 2, or consider all other points in the dataset), do the following:

[0189] Get the k-distance of data point o and calculate the actual distance from data point p to its neighbor o; store the calculated reachable distance in an appropriate data structure for subsequent calculation of local reachability density and local anomaly factor score.

[0190] In another preferred embodiment of the present invention, the above step 15 may include:

[0191] For each data point p in the working state data, initialize a variable to store its local reachable density (LRD); for each data point p, obtain the list of its k nearest neighbors (calculated in the previous step); traverse each neighbor in this neighbor list o i ; For each neighbor o i , get the data point p to o i The reachable distances of data point p (calculated in the previous step) are summed up. The sum of the accumulated reachable distances is divided by k (the number of neighbors) to get the average reachable distance. The local reachability density (LRD) of data point p is defined as the inverse of this average reachable distance.

[0192] For each data point p, initialize a variable to store the average local reachability density of its k nearest neighbors, and traverse the k nearest neighbors of the data point p again:

[0193] Iterate through each neighbor in the k nearest neighbor list of data point p o i ; Accumulate the local reachability density of neighbors:

[0194] For each neighbor o i , obtain its local reachability density (calculated in the above steps), and add up these local reachability densities. Divide the sum of the accumulated local reachability densities by k to get the average local reachability density of the k nearest neighbors of the data point p.

[0195] For each data point p, initialize a variable to store its local outlier factor score (LOF), and calculate the LOF score:

[0196] The local outlier factor score (LOF) of data point p is obtained by dividing the average local reachability density of the k nearest neighbors of data point p by the local reachability density of data point p itself. The LOF score reflects the degree of outlier of data point p relative to its neighbors: the higher the score, the greater the density difference between p and its neighbors, and therefore the more likely p is an outlier.

[0197] For example, detecting abnormal points in the working status data of a vacuum pump includes:

[0198] Suppose there is a data set of the working status of a vacuum pump, which contains data in multiple dimensions such as the pump's operating time, temperature, and pressure. For simplicity, only two-dimensional data is considered: operating time and pressure.

[0199] Select k=3, that is, consider the nearest 3 neighbors of each data point; for each point in the data set, use the Euclidean distance to find the nearest 3 neighbors. For each data point, calculate its distance to the third nearest neighbor as the k-distance. Then, for each neighbor of the point, calculate the reachable distance, that is, the larger value of the actual distance from the data point to the neighbor and the k-distance of the neighbor. For each point in the data set, initialize three variables: local reachable density (LRD), the average local reachable density of the k nearest neighbors, and the local outlier factor score (LOF). For each data point, accumulate its reachable distance to its 3 nearest neighbors and divide it by 3 to get the average reachable distance. Then, take the inverse of this average reachable distance as the local reachable density of the point. For each data point, traverse its 3 nearest neighbors again, accumulate their local reachable densities, and divide it by 3 to get the average local reachable density. Finally, divide the average local reachable density of the k nearest neighbors of each data point by the local reachable density of the point to get the local outlier factor score of the point.

[0200] In another preferred embodiment of the present invention, the above step 16 may include:

[0201] According to historical data, a reasonable local anomaly factor (LOF) score threshold is set. This threshold is used to determine which data points are considered abnormal. This threshold is set slightly higher than the LOF score of normal data points to ensure that only true anomalies trigger warnings.

[0202] During the operation of the vacuum pump, the working status data is collected in real time, and the LOF score of each new data point is calculated according to the method of step 14 and step 15. This can be achieved by adding the new data point to the existing data set and recalculating the affected local area (such as the k nearest neighbors of the new data point).

[0203] The LOF score of the new data point is compared with the preset warning threshold. If the LOF score exceeds the threshold, it indicates that the data point is an abnormal point, which may mean that the vacuum pump is in an abnormal working state.

[0204] Once an abnormal point is detected, the system should immediately trigger an early warning mechanism. This can be achieved in a variety of ways, such as sending an email or SMS notification to relevant personnel, or displaying a clear warning message on the monitoring interface. The early warning information should contain detailed information about the abnormal point, such as the time of occurrence, abnormal parameter value, and possible abnormal situation description.

[0205] When triggering an early warning, the system should also automatically record relevant information about the abnormal point, including data value, timestamp, and LOF score. This information is crucial for subsequent troubleshooting, cause analysis, and formulation of preventive measures. In addition, through the analysis of historical abnormal data, it can also help optimize the setting of early warning thresholds and improve the accuracy of anomaly detection.

[0206] After receiving the early warning, relevant personnel should respond quickly and take measures to deal with it. This may include checking the operating status of the vacuum pump, adjusting operating parameters, performing necessary maintenance or replacing damaged parts, etc. Through timely response and disposal, the impact of abnormal conditions on vacuum pump performance and production processes can be minimized.

[0207] like Figure 2 As shown, an embodiment of the present invention further provides a vacuum pump real-time data anomaly detection system 20, comprising:

[0208] The acquisition module 21 is used to collect the working status data of the vacuum pump in real time, including vacuum degree, temperature, pressure, and vibration frequency;

[0209] The sending module 22 is used to transmit the working status data to the data processor in real time via a data line or a wireless network;

[0210] The analysis module 23 is used to analyze the physical phenomena of the vacuum pump in the current working state according to the working state data to obtain analysis results, where the physical phenomena include stress distribution, temperature change and pressure fluctuation;

[0211] The first calculation module 24 is used to find k nearest neighbors for each data point p in the working status data according to the analysis results, sort these neighbors from near to far according to the distance, calculate the k-distance, and record the distance from the data point p to its kth nearest neighbor as the k-distance of p; calculate the reachable distance from p to each neighbor o, for each neighbor o of the data point p, obtain the k-distance of the data point o, and calculate the actual distance from the data point p to the neighbor o; the reachable distance from the data point p to the data point o is defined as the maximum value between the k-distance of the data point o and the actual distance from the data point p to the data point o;

[0212] The second calculation module 25 is used to calculate the inverse of the sum of the reachable distances of its k nearest neighbors for each data point p to obtain the local reachable density; for each data point p, calculate the average value of the local reachable densities of its k nearest neighbors; divide the average value by the local reachable density of the data point p itself to obtain the local anomaly factor score of the data point p;

[0213] The detection module 26 is used to trigger an early warning mechanism when an abnormal situation is detected based on the local abnormal factor score.

[0214] It should be noted that the system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0215] The embodiment of the present invention further provides a computing device, comprising: a processor, a memory storing a computer program, wherein when the computer program is executed by the processor, the method described above is executed. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0216] The embodiment of the present invention also provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described above. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0217] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for detecting abnormalities in real-time data of a vacuum pump, characterized in that: The method comprises: Real-time collection of vacuum pump working status data, including vacuum degree, temperature, pressure, and vibration frequency; Transmit working status data to the data processor in real time via data line or wireless network; According to the working status data, the physical phenomena of the vacuum pump in the current working state are analyzed to obtain the analysis results. The physical phenomena include stress distribution, temperature change and pressure fluctuation, including: extracting the vibration frequency data related to stress from the working status data; establishing the mechanical model of the vacuum pump according to the structural design of the vacuum pump; calculating the stress distribution of the vacuum pump in the current working state according to the vibration frequency data and the mechanical model of the vacuum pump to obtain the stress distribution analysis results; extracting the temperature data from the working status data, and drawing the temperature curve of the temperature change over time according to the temperature data and the corresponding timestamp; detecting the temperature curve through the preset temperature threshold. Abnormal points; analyze the thermal performance of the vacuum pump during operation according to the temperature change curve and abnormal points to obtain the temperature change analysis results; extract pressure data from the working status data; calculate the coefficient of variation of the pressure data and draw a pressure fluctuation curve of the pressure changing with time; detect abnormal fluctuations in the pressure curve through a preset fluctuation threshold; analyze the cause of abnormal fluctuations according to the pressure fluctuation curve and abnormal fluctuation points to obtain the pressure fluctuation analysis results; conduct a comprehensive analysis of the stress distribution analysis results, temperature change analysis results and pressure fluctuation analysis results to obtain a comprehensive analysis result, which includes the existing failure modes, performance degradation trends or safety hazards; According to the analysis results, for each data point p in the working status data, find the k nearest neighbors, sort these neighbors from near to far, calculate the k-distance, and record the distance from the data point p to its kth nearest neighbor as the k-distance of p; calculate the reachable distance from p to each neighbor o, for each neighbor o of the data point p, obtain the k-distance of the data point o, and calculate the actual distance from the data point p to the neighbor o; the reachable distance from the data point p to the data point o is defined as the maximum value between the k-distance of the data point o and the actual distance from the data point p to the data point o; For each data point p, the local reachability density is obtained by calculating the inverse of the sum of the reachability distances of its k nearest neighbors; for each data point p, the average of the local reachability densities of its k nearest neighbors is calculated; the average is divided by the local reachability density of the data point p itself to obtain the local anomaly factor score of the data point p; According to the local anomaly factor score, when an abnormal situation is detected, the early warning mechanism is triggered.

2. A vacuum pump real-time data anomaly detection method according to claim 1, characterized in that: The working status data is transmitted to the data processor in real time via data line or wireless network, including: Process the working status data and extract key features, including switch status, operating time, vacuum degree, pumping rate, ambient temperature, humidity, lubricating oil status, filter cleanliness, abnormal vibration and high temperature alarm; According to key features, the working status data is divided into different categories to obtain classification results. The different categories include operation data, performance data, environmental data, maintenance data, and fault and warning data; According to the classification results, each category of data is packaged into different data packets, and corresponding metadata is attached to each data packet, including timestamp and data source information; For each data packet, the urgency of each data packet is evaluated according to the data category and specific content contained therein, so as to obtain an urgency evaluation result; According to the evaluation result of urgency, the transmission priority of each data packet is set; According to the priority of the data packet, the data packet is transmitted to the data processor in real time through the data line or wireless network.

3. A vacuum pump real-time data anomaly detection method according to claim 2, characterized in that: According to key features, the work status data is divided into different categories to obtain classification results, including: Encode the key features and randomly generate an initial population, where each individual in the initial population represents a combination of key features; Define a fitness function to evaluate the contribution of each key feature combination to classification performance; According to the fitness function, the corresponding individuals are determined to enter the next generation, and two individuals are randomly selected for crossover operation to generate a new combination of key features; The gene positions of individuals are mutated with a certain probability, and the genetic operation is repeated until the preset number of iterations is reached to obtain the final combination of key features; According to the preset classifier and the final combination of key features, the working status data is divided into different categories to obtain the classification results.

4. A vacuum pump real-time data anomaly detection method according to claim 3, characterized in that: Encode the key features and randomly generate an initial population, including: Determine a list of key characteristics, including switch status, operating hours, vacuum level, pumping rate, ambient temperature, humidity, lubricant condition, filter cleanliness, abnormal vibration and high temperature alarm; According to the key feature list, determine the number of key features and the encoding length of each key feature; According to the encoding length of each key feature, the selection state of each key feature is converted into a corresponding binary code; Set the number of individuals in the population, and for each individual, randomly generate a binary string equal to the number of key features; Combine all randomly generated individuals into an initial population.

5. A vacuum pump real-time data anomaly detection method according to claim 4, characterized in that: For each data packet, the urgency of each data packet is evaluated according to the data category and specific content contained therein to obtain an urgency evaluation result, including: Analyze the nature and importance of each data item in each data packet and assign a weight to each data item; Assign an urgency level to each data item based on its time sensitivity and immediacy factors; For each data item, multiply its weight by the value of its urgency level to obtain the urgency score of the data item; The urgency scores of all data items are added together to obtain the urgency evaluation result of the entire data packet.

6. A vacuum pump real-time data anomaly detection method according to claim 5, characterized in that: Analyze the nature and importance of each data item in each data packet and assign a weight to each data item, including: Classify all data items according to the physical quantities or states they reflect to obtain classification results; For each data item, analyze the properties it represents and evaluate the importance of each data item to the vacuum pump operating status monitoring; A weight is assigned to each data item according to its importance to the monitoring of the operating status of the vacuum pump.

7. A vacuum pump real-time data anomaly detection system, characterized in that: Applied to the method according to any one of claims 1 to 6, comprising: The acquisition module is used to collect the working status data of the vacuum pump in real time, including vacuum degree, temperature, pressure, and vibration frequency; A sending module, used for transmitting the working status data to the data processor in real time via a data line or a wireless network; An analysis module is used to analyze the physical phenomena of the vacuum pump in the current working state according to the working state data to obtain analysis results, the physical phenomena including stress distribution, temperature change and pressure fluctuation; The first calculation module is used to find k nearest neighbors for each data point p in the working status data according to the analysis results, sort these neighbors from near to far according to the distance, calculate the k-distance, and record the distance from the data point p to its kth nearest neighbor as the k-distance of p; calculate the reachable distance from p to each neighbor o, for each neighbor o of the data point p, obtain the k-distance of the data point o, and calculate the actual distance from the data point p to the neighbor o; the reachable distance from the data point p to the data point o is defined as the maximum value between the k-distance of the data point o and the actual distance from the data point p to the data point o; The second calculation module is used to calculate the inverse of the sum of the reachable distances of its k nearest neighbors for each data point p to obtain the local reachable density; for each data point p, calculate the average value of the local reachable densities of its k nearest neighbors; divide the average value by the local reachable density of the data point p itself to obtain the local anomaly factor score of the data point p; The detection module is used to trigger the early warning mechanism when an abnormal situation is detected based on the local abnormal factor score.

8. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images