A contact life prediction method based on big data

Abstract

The present invention discloses a big data-based contact life prediction method, which relates to the technical field of contact life prediction. The method collects contact operating data, performs multimodal feature fusion on the operating data, extracts the arc energy main frequency, impact damage amount, temperature rise rate, and environmental corrosion index, and constructs an N×8-dimensional feature matrix. The method then calculates the final predicted life data based on step S2, obtains the vibration correction factor and resistance change rate in real time, and verifies the final predicted life data and optimizes model parameters using a digital twin platform. The method establishes a mechanical correction factor by comprehensively analyzing various contact operating data, particularly the frequency response of vibration frequency and impact force. Combined with contact stability, the method significantly improves the accuracy of contact life prediction. Furthermore, the method enables real-time prediction and early warning, enhances adaptability, reduces equipment failure rate and maintenance costs, and provides a strong guarantee for the safe and stable operation of power systems.

Description

Technical Field

[0001] The present invention relates to the technical field of contact life prediction, and in particular to a contact life prediction method based on big data. Background Art

[0002] In power systems, contacts are critical components of electrical equipment, and their performance and lifespan are directly related to the stability and safety of the entire system. With the continuous development of power systems, the requirements for contact performance are becoming increasingly stringent. However, during operation, contacts are affected by a variety of factors, such as arc discharge, mechanical vibration, temperature fluctuations, and environmental corrosion. These factors can cause contact performance to gradually deteriorate and eventually fail. Traditional contact life prediction methods often rely on empirical formulas or periodic testing. These methods not only have limited prediction accuracy but also fail to reflect the actual operating status of the contacts in a timely manner.

[0003] At present, the Chinese patent application number CN202211104655.4 discloses a life prediction method, device, circuit breaker and medium for circuit breaker contacts. The method includes: obtaining the impedance correction coefficient of the contact group in the circuit breaker; obtaining the current thermal capacity of the contact group based on the impedance correction coefficient and the breaking current of the circuit breaker, where the breaking current is the current of the closed loop before the circuit breaker performs the breaking operation; determining the remaining life parameters of the contact group based on the initial thermal capacity and current thermal capacity of the contact group. This realizes that the remaining life parameters of the contact group can be determined based on the impedance correction coefficient of the contact group and the initial thermal capacity and current thermal capacity of the contact group. Compared with the existing technology, there is no need to make major changes to the structural design of the circuit breaker, which can reduce the prediction cost of the life prediction method.

[0004] Related technologies make it difficult to predict lifespan by analyzing the frequency response of vibration frequency and impact force combined with the stability of contact, which affects the accuracy of lifespan prediction. Summary of the Invention

[0005] The technical problem solved by the present invention is that it is difficult to predict the lifespan by analyzing the frequency response of vibration frequency and impact force in combination with the stability of contact, which affects the accuracy of lifespan prediction.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A contact life prediction method based on big data includes the following steps:

[0008] Step S1, collecting contact operation data;

[0009] Step S2: Multimodal feature fusion is performed on the operating data to extract the arc energy main frequency, impact damage amount, temperature rise rate and environmental corrosion index and construct dimensional feature matrix;

[0010] Step S3, calculating the final predicted lifespan data based on step S2;

[0011] Step S4, obtaining the vibration correction factor and resistance change rate in real time;

[0012] Step S5: Verify the final predicted life data and optimize the model parameters in combination with the digital twin platform.

[0013] Preferably, step S1 includes the following sub-steps:

[0014] Step S101: deploying high-frequency sensors to collect electrical parameters, including real-time current data, real-time voltage data, and real-time arc energy data;

[0015] Step S102 , collecting mechanical parameters using a triaxial accelerometer, the mechanical parameters including vibration acceleration data, impact force spectrum data, and contact pressure fluctuation data;

[0016] Step S103, using an infrared thermal imager to monitor and obtain contact surface parameters, wherein the contact surface parameters include a contact surface temperature rise curve and contact surface heat dissipation rate data;

[0017] Step S104, collecting environmental data through environmental sensors, wherein the environmental data includes temperature data, humidity data, salt spray concentration data, and corrosive gas concentration data;

[0018] Step S105 : combining the electrical parameters, mechanical parameters, contact surface parameters, and environmental parameters to output as operating data.

[0019] Preferably, step S2 includes the following sub-steps:

[0020] Step S201: extracting arc energy characteristics based on electrical parameters. The arc energy characteristics include the cumulative value of arc energy. The mathematical expression of the cumulative value of arc energy is:

[0021] ;

[0022] in, is the accumulated value of arc energy, For real-time current data, For real-time voltage data, Arc duration data in real-time arc energy data;

[0023] Extract the frequency domain distribution characteristics of real-time arc energy data, analyze the main frequency component of the arc energy data through fast Fourier transform, and output the main frequency of the arc energy;

[0024] Step S202, performing wavelet packet decomposition on the vibration acceleration data to obtain energy entropy of each frequency band of the vibration acceleration data, and outputting the energy entropy as vibration energy entropy;

[0025] The peak frequency and attenuation coefficient of the impact force spectrum data are analyzed to construct the impact damage equivalent. The mathematical expression of the impact damage equivalent is:

[0026] ;

[0027] in, is the impact damage equivalent, is the initial peak value of the impact force spectrum data, is the attenuation coefficient, is the contact quality, is the total impact time, is the time from the start of the impact; is the natural frequency of the contact structure, is the peak frequency.

[0028] Preferably, the step S2 further includes the following sub-steps:

[0029] Step S203, fitting the contact surface temperature rise curve to obtain the temperature rise rate and exponential factor;

[0030] Step S204: Calculate the environmental corrosion index based on the salt spray concentration data and the humidity data. The mathematical expression of the environmental corrosion index is:

[0031] ;

[0032] in, is the environmental corrosion index, is the salt spray concentration data, is the humidity data, is the preset humidity threshold data, is the saturated humidity data;

[0033] Step S205: normalize the arc energy main frequency, impact damage amount, temperature rise rate and environmental corrosion index to construct dimensional feature matrix.

[0034] Preferably, step S3 includes the following sub-steps:

[0035] Step S301: Calculate contact erosion data based on real-time arc energy data. The mathematical expression of the contact erosion data is:

[0036] ;

[0037] in, is the contact ablation data, is the maximum temperature rise of the contact surface temperature rise curve, is the preset ablation rate constant, is the preset temperature sensitivity coefficient;

[0038] An oxidation life benchmark model is established based on the Arrhenius equation. The oxidation life benchmark model is:

[0039] ;

[0040] in, is the baseline prediction value of oxidation life, is the preset pre-exponential factor, is the activation energy, is the gas constant, is the temperature data;

[0041] Step S302: Calculate a mechanical correction factor based on the impact damage equivalent and the vibration energy entropy. The mathematical expression of the mechanical correction factor is:

[0042] ;

[0043] in, is the mechanical correction factor, is the vibration energy entropy, is the impact damage equivalent threshold.

[0044] Preferably, the step S3 further includes the following sub-steps:

[0045] Step S303, correcting the oxidation life by the environmental corrosion index. The mathematical expression of the corrected oxidation life is:

[0046] ;

[0047] in, To correct for oxidation lifetime, is the critical corrosion index.

[0048] Preferably, the step S3 further includes the following sub-steps:

[0049] Step S304: The dimensional feature matrix is ​​input into the attention mechanism long short-term memory neural network to obtain the predicted remaining life data;

[0050] The predicted remaining life data is dynamically weighted and fused. The mathematical expression of the dynamic weighted fusion is:

[0051] ;

[0052] in, For the final predicted lifespan data, is the critical ablation amount;

[0053] Output the final predicted life data.

[0054] Preferably, the step S4 includes the following sub-steps:

[0055] Step S401: deploy a lightweight model at the edge to obtain the vibration correction factor and resistance change rate in real time;

[0056] In step S402 , the cloud receives edge feature data and generates a remaining life probability density function based on Monte Carlo simulation.

[0057] Preferably, the logic of the digital twin verification method in step S5 is:

[0058] Microcrack extension data, oxide layer thickening data and arc erosion data are collected, weights are set for the microcrack extension data, oxide layer thickening data and arc erosion data respectively, and the reward function is obtained. The reward function is input as reinforcement learning input data into the attention mechanism long short-term memory neural network.

[0059] Preferably, the method further includes step S6, wherein when the salt mist concentration data is greater than a preset salt mist concentration threshold, the method switches to a preset corrosion-dominated prediction model.

[0060] The present invention achieves beneficial effects by comprehensively analyzing various contact operating data, particularly the vibration frequency and impact force frequency response, to establish a mechanical correction factor. Combined with contact stability, this significantly improves the accuracy of contact life prediction. Furthermore, it enables real-time prediction and early warning, enhances adaptability, reduces equipment failure rates and maintenance costs, and provides a strong guarantee for the safe and stable operation of power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A flowchart of the steps of a contact life prediction method based on big data is provided in accordance with one embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0063] Example, see Figure 1 , provides a contact life prediction method based on big data, including the following steps:

[0064] Step S1: collecting operating data of the contacts.

[0065] Step S2: Multimodal feature fusion is performed on the operating data to extract the arc energy main frequency, impact damage amount, temperature rise rate and environmental corrosion index and construct dimensional feature matrix.

[0066] Step S3, calculating the final predicted life data based on step S2.

[0067] Step S4: obtaining the vibration correction factor and resistance change rate in real time.

[0068] Step S5: Verify the final predicted life data and optimize the model parameters in combination with the digital twin platform.

[0069] Step S1 includes the following sub-steps:

[0070] Step S101 : deploying high-frequency sensors to collect electrical parameters, including real-time current data, real-time voltage data, and real-time arc energy data.

[0071] High-frequency sensors can collect contact current, voltage and arc energy data in real time and accurately, reflecting the electrical operating status of the contacts and providing basic data for analyzing the arc discharge characteristics and electrical life of the contacts.

[0072] Step S102 : collecting mechanical parameters through a three-axis accelerometer, where the mechanical parameters include vibration acceleration data, impact force spectrum data, and contact pressure fluctuation data.

[0073] The triaxial accelerometer can capture the dynamic response of the contact under mechanical vibration and impact force, and evaluate the mechanical stress and damage of the contact through vibration acceleration data, impact force spectrum data and contact pressure fluctuation data.

[0074] Step S103 : Using an infrared thermal imager to monitor and obtain contact surface parameters, the contact surface parameters include a contact surface temperature rise curve and contact surface heat dissipation rate data.

[0075] Infrared thermal imagers can monitor the temperature distribution and heat dissipation rate on the contact surface in real time, reflecting the temperature changes of the contact under heat load, and providing an important basis for analyzing the thermal life and heat dissipation performance of the contact.

[0076] Step S104 , collecting environmental data through environmental sensors, the environmental data including temperature data, humidity data, salt spray concentration data, and corrosive gas concentration data.

[0077] Environmental sensors can collect real-time environmental data of the contacts, including temperature, humidity, salt spray concentration, and corrosive gas concentration. These environmental data are of great significance for evaluating the environmental corrosion effects of contacts and predicting contact life.

[0078] Step S105 : combining the electrical parameters, mechanical parameters, contact surface parameters, and environmental parameters to output as operating data.

[0079] The above-mentioned parameters are combined and output as operating data, providing a comprehensive and unified data basis for subsequent multimodal feature fusion and contact life prediction, ensuring the accuracy and reliability of the prediction results.

[0080] Step S1 deploys a variety of sensors to comprehensively collect the electrical parameters, mechanical parameters, contact surface parameters and environmental parameters of the contact during operation, providing comprehensive and accurate data support for subsequent multimodal feature fusion and contact life prediction.

[0081] Step S2 includes the following sub-steps:

[0082] Step S201: extract arc energy characteristics based on electrical parameters. The arc energy characteristics include the cumulative value of arc energy. The mathematical expression of the cumulative value of arc energy is:

[0083] ;

[0084] in, is the accumulated value of arc energy, For real-time current data, For real-time voltage data, It is the arc duration data in the real-time arc energy data.

[0085] The frequency domain distribution characteristics of real-time arc energy data are extracted, and the main frequency component of the arc energy data is analyzed by fast Fourier transform, and the output is the main frequency of the arc energy.

[0086] Step S201 extracts the cumulative value of arc energy, which reflects the energy consumption of the contact during the arc discharge process. The main frequency component of the arc energy data is analyzed by fast Fourier transform to obtain the main frequency of the arc energy, which helps to identify the characteristics and stability of the arc discharge.

[0087] Step S202 : performing wavelet packet decomposition on the vibration acceleration data to obtain energy entropy of each frequency band of the vibration acceleration data, and outputting the energy entropy as vibration energy entropy.

[0088] The peak frequency and attenuation coefficient of the impact force spectrum data are analyzed to construct the impact damage equivalent. The mathematical expression of the impact damage equivalent is:

[0089] ;

[0090] in, is the impact damage equivalent, is the initial peak value of the impact force spectrum data, is the attenuation coefficient, is the contact quality, is the total impact time, is the time from the start of the impact; is the natural frequency of the contact structure, is the peak frequency.

[0091] Step S202 performs wavelet packet decomposition on the vibration acceleration data to obtain the energy entropy of each frequency band, namely, the vibration energy entropy, which reflects the energy distribution and dissipation of the contact under the action of vibration. The peak frequency and attenuation coefficient of the impact force spectrum data are analyzed to construct the impact damage equivalent, which helps to evaluate the degree of damage caused by the impact force to the contact.

[0092] Step S2 also includes the following sub-steps:

[0093] Step S203: Fitting the contact surface temperature rise curve to obtain the temperature rise rate and exponential factor.

[0094] Step S203 fits the contact surface temperature rise curve to obtain the temperature rise rate and exponential factor, which reflect the temperature change law and heat dissipation performance of the contact under the action of heat load.

[0095] Step S204: Calculate the environmental corrosion index based on the salt spray concentration data and the humidity data. The mathematical expression of the environmental corrosion index is:

[0096] ;

[0097] in, is the environmental corrosion index, is the salt spray concentration data, is the humidity data, is the preset humidity threshold data, The saturated humidity data.

[0098] Step S204 calculates the environmental corrosion index based on the salt spray concentration data and the humidity data, which helps to evaluate the corrosiveness of the environment in which the contacts are located and its impact on the life of the contacts.

[0099] Step S205: normalize the arc energy main frequency, impact damage amount, temperature rise rate and environmental corrosion index to construct dimensional feature matrix.

[0100] Step S205 normalizes the features extracted above and constructs dimensional feature matrix, which ensures the weight balance of different features in the subsequent prediction model and improves the accuracy of the prediction. At the same time, the construction of the feature matrix also provides a unified data format and input interface for subsequent deep learning models.

[0101] Step S2 constructs a deep feature extraction model of the collected operation data, including arc energy characteristics, vibration and impact characteristics, contact surface temperature rise characteristics, and environmental corrosion characteristics. The dimensional feature matrix provides key feature input for subsequent contact life prediction. These features comprehensively reflect the electrical, mechanical, thermal and environmental status information of the contact during operation.

[0102] Step S3 includes the following sub-steps:

[0103] Step S301: Calculate contact erosion data based on real-time arc energy data. The mathematical expression of the contact erosion data is:

[0104] ;

[0105] in, is the contact ablation data, is the maximum temperature rise of the contact surface temperature rise curve, is the preset ablation rate constant, is the preset temperature sensitivity coefficient.

[0106] An oxidation life benchmark model is established based on the Arrhenius equation. The oxidation life benchmark model is:

[0107] ;

[0108] in, is the baseline prediction value of oxidation life, is the preset pre-exponential factor, is the activation energy, is the gas constant, is the temperature data.

[0109] The contact erosion data is calculated based on the real-time arc energy data, which helps to directly reflect the degree of erosion of the contact material by the arc discharge. The oxidation life benchmark model is established through the Arrhenius equation, which provides a theoretical basis for evaluating the life of the contact under oxidation.

[0110] Step S302: Calculate the mechanical correction factor based on the impact damage equivalent and the vibration energy entropy. The mathematical expression of the mechanical correction factor is:

[0111] ;

[0112] in, is the mechanical correction factor, is the vibration energy entropy, is the impact damage equivalent threshold.

[0113] The mechanical correction factor is calculated based on the impact damage equivalent and vibration energy entropy, which takes into account the additional impact of mechanical vibration and impact force on the contact life and improves the accuracy of life prediction.

[0114] Step S3 also includes the following sub-steps:

[0115] Step S303, correcting the oxidation life by the environmental corrosion index. The mathematical expression of the corrected oxidation life is:

[0116] ;

[0117] in, To correct for oxidation lifetime, is the critical corrosion index.

[0118] The oxidation life is corrected by the environmental corrosion index, which reflects the potential impact of environmental factors on contact life and makes the prediction results more consistent with actual conditions.

[0119] Step S3 also includes the following sub-steps:

[0120] Step S304: The dimensional feature matrix is ​​input into the attention mechanism long short-term memory neural network to obtain the predicted remaining life data.

[0121] The predicted remaining life data is dynamically weighted and fused. The mathematical expression of dynamic weighted fusion is:

[0122] ;

[0123] in, For the final predicted lifespan data, is the critical ablation amount.

[0124] Output the final predicted life data.

[0125] Will The dimensional feature matrix is ​​input into the attention mechanism long short-term memory neural network, which fully utilizes the advantages of deep learning models in processing complex data relationships and improves the accuracy of prediction.

[0126] The predicted remaining life data is dynamically weighted and fused, which takes into account the relative importance of different prediction results and further improves the reliability of the final predicted life data.

[0127] Step S3 comprehensively applies real-time arc energy data, mechanical parameters, environmental corrosion index, and an N*8-dimensional feature matrix. By calculating the contact ablation amount and mechanical correction factor, and considering the impact of environmental corrosion on oxidation life, the remaining life of the contacts is predicted using a long short-term memory neural network with an attention mechanism. This process not only considers the comprehensive impact of electrical, mechanical, and environmental factors on the contacts, but also improves the accuracy of predictions through deep learning models, providing strong support for contact maintenance and management.

[0128] Step S4 includes the following sub-steps:

[0129] Step S401: deploy a lightweight model at the edge to obtain the vibration correction factor and resistance change rate in real time.

[0130] Deploying lightweight models at the edge reduces computing resource consumption, enabling real-time acquisition of key parameters such as vibration correction factors and resistance change rates even in resource-constrained environments. This not only improves the real-time nature of data acquisition but also helps reduce data transmission latency and costs.

[0131] Real-time acquisition of the vibration correction factor and resistance change rate are crucial for assessing the operating status of equipment and predicting its remaining lifespan. The vibration correction factor reflects changes in equipment performance under vibration, while the resistance change rate reflects changes in the device's internal resistance, indirectly reflecting the wear and aging of the device.

[0132] In step S402 , the cloud receives edge feature data and generates a remaining life probability density function based on Monte Carlo simulation.

[0133] The cloud receives edge feature data. This step realizes the aggregation and centralized processing of data, providing a reliable data source for subsequent calculations and simulations.

[0134] Monte Carlo simulation is used to generate a probability density function (PDF) for the remaining life of a device. This method simulates multiple possible equipment operating states and lifespan scenarios to generate a probability density function that reflects the distribution of the remaining life of the device. This not only provides more comprehensive lifespan prediction information, but also helps users understand the uncertainty of equipment lifespan, providing a more scientific basis for equipment maintenance decisions.

[0135] Step S4 enables collaboration between the edge and the cloud to efficiently and in real time assess and predict the remaining life of the equipment. By deploying a lightweight model at the edge, key operating parameters can be quickly acquired, providing the necessary data support for complex calculations and simulations in the cloud. The cloud, using the received edge feature data, generates a probability density function for the remaining life through Monte Carlo simulation. This provides a more comprehensive understanding of the distribution of equipment lifespan, providing a scientific basis for equipment maintenance, replacement, and fault warning.

[0136] The logic of the digital twin verification method in step S5 is:

[0137] Microcrack extension data, oxide layer thickening data and arc erosion data are collected, weights are set for the microcrack extension data, oxide layer thickening data and arc erosion data respectively, and the reward function is obtained. The reward function is input as reinforcement learning input data into the attention mechanism long short-term memory neural network.

[0138] Step S5 collects data on microcrack growth, oxide layer thickening, and arc erosion, and assigns weights to each of these data to construct a reward function that comprehensively reflects the actual operating status of the contact. This reward function is then fed into the attention mechanism long-short-term memory neural network as input for reinforcement learning. This process fully leverages the advantages of digital twin technology, simulating and verifying the actual operating status of the contacts, providing a more accurate and reliable basis for contact life prediction. Furthermore, the introduction of the attention mechanism long-short-term memory neural network further improves the accuracy and generalization of the prediction model, providing strong support for contact maintenance and management.

[0139] The method further includes step S6, when the salt mist concentration data is greater than a preset salt mist concentration threshold, switching to a preset corrosion-dominant prediction model.

[0140] The corrosion-dominant prediction model is optimized for corrosive environments, enabling more accurate predictions of contact life in such environments. Through intelligent switching, the system can provide even more reliable predictions in conditions with high salt spray concentrations.

[0141] This method collects various operating data of the contacts and constructs The dimensional characteristic matrix comprehensively considers various factors that may affect the contact during operation, thereby improving prediction accuracy. In particular, by analyzing the frequency response of vibration frequency and impact force, it can more accurately reflect the damage of the contact under mechanical stress, providing more reliable data support for life prediction.

[0142] By analyzing the frequency response of vibration and impact forces, this method establishes a mechanical correction factor. This correction factor reflects changes in contact performance under mechanical stress, enabling a more accurate prediction of the remaining contact life. Furthermore, by incorporating contact stability, the prediction results can be further refined, improving both accuracy and reliability.

[0143] This method deploys a lightweight model at the edge to obtain vibration correction factors and resistance change rates in real time. Monte Carlo simulation is then used in the cloud to generate a probability density function for the remaining lifespan, enabling real-time prediction and early warning of contact lifespan. This not only enables timely detection of potential contact failures but also provides a scientific basis for equipment maintenance and management, reducing equipment failure rates and repair costs.

[0144] This method also considers the impact of environmental corrosion on contact life. When the salt spray concentration data exceeds a preset threshold, it automatically switches to a preset corrosion-dominated prediction model. This design enables the method to adapt to the contact life prediction needs in different environments, improving its applicability and flexibility.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0146] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A contact life prediction method based on big data, characterized in that: The steps include: Step S1, collecting contact operation data; Step S2: Multimodal feature fusion is performed on the operating data to extract the arc energy main frequency, impact damage amount, temperature rise rate and environmental corrosion index and construct dimensional feature matrix; Step S3, calculating the final predicted lifespan data based on step S2; Step S4, obtaining the vibration correction factor and resistance change rate in real time; Step S5: Verify the final predicted lifespan data and optimize model parameters in conjunction with the digital twin platform; The step S3 includes the following sub-steps: Step S301: Calculate contact erosion data based on real-time arc energy data. The mathematical expression of the contact erosion data is: ； in, is the contact ablation data, is the maximum temperature rise of the contact surface temperature rise curve, is the preset ablation rate constant, is the preset temperature sensitivity coefficient; An oxidation life benchmark model is established based on the Arrhenius equation. The oxidation life benchmark model is: ； in, is the baseline prediction value of oxidation life, is the preset pre-exponential factor, is the activation energy, is the gas constant, is the temperature data; Step S302: Calculate a mechanical correction factor based on the impact damage equivalent and the vibration energy entropy. The mathematical expression of the mechanical correction factor is: ； in, is the mechanical correction factor, is the vibration energy entropy, is the impact damage equivalent threshold; The step S3 further includes the following sub-steps: Step S303, correcting the oxidation life by the environmental corrosion index. The mathematical expression of the corrected oxidation life is: ； in, To correct for oxidation lifetime, is the critical corrosion index; The step S3 further includes the following sub-steps: Step S304: The dimensional feature matrix is ​​input into the attention mechanism long short-term memory neural network to obtain the predicted remaining life data; The predicted remaining life data is dynamically weighted and fused. The mathematical expression of the dynamic weighted fusion is: ； in, For the final predicted lifespan data, is the critical ablation amount; Output the final predicted life data.

2. The contact life prediction method based on big data according to claim 1, characterized in that: The step S1 includes the following sub-steps: Step S101: deploying high-frequency sensors to collect electrical parameters, including real-time current data, real-time voltage data, and real-time arc energy data; Step S102 , collecting mechanical parameters using a triaxial accelerometer, the mechanical parameters including vibration acceleration data, impact force spectrum data, and contact pressure fluctuation data; Step S103, using an infrared thermal imager to monitor and obtain contact surface parameters, wherein the contact surface parameters include a contact surface temperature rise curve and contact surface heat dissipation rate data; Step S104, collecting environmental data through environmental sensors, wherein the environmental data includes temperature data, humidity data, salt spray concentration data, and corrosive gas concentration data; Step S105 : combining the electrical parameters, mechanical parameters, contact surface parameters, and environmental parameters to output as operating data.

3. The contact life prediction method based on big data according to claim 2, characterized in that: The step S2 includes the following sub-steps: Step S201: extracting arc energy characteristics based on electrical parameters. The arc energy characteristics include the cumulative value of arc energy. The mathematical expression of the cumulative value of arc energy is: ； in, is the accumulated value of arc energy, For real-time current data, For real-time voltage data, Arc duration data in real-time arc energy data; Extract the frequency domain distribution characteristics of real-time arc energy data, analyze the main frequency component of the arc energy data through fast Fourier transform, and output the main frequency of the arc energy; Step S202, performing wavelet packet decomposition on the vibration acceleration data to obtain energy entropy of each frequency band of the vibration acceleration data, and outputting the energy entropy as vibration energy entropy; The peak frequency and attenuation coefficient of the impact force spectrum data are analyzed to construct the impact damage equivalent. The mathematical expression of the impact damage equivalent is: ； in, is the impact damage equivalent, is the initial peak value of the impact force spectrum data, is the attenuation coefficient, is the contact quality, is the total impact time, is the time from the start of the impact; is the natural frequency of the contact structure, is the peak frequency.

4. The contact life prediction method based on big data according to claim 3, characterized in that: The step S2 further includes the following sub-steps: Step S203, fitting the contact surface temperature rise curve to obtain the temperature rise rate and exponential factor; Step S204: Calculate the environmental corrosion index based on the salt spray concentration data and the humidity data. The mathematical expression of the environmental corrosion index is: ； in, is the environmental corrosion index, is the salt spray concentration data, is the humidity data, is the preset humidity threshold data, is the saturated humidity data; Step S205: normalize the arc energy main frequency, impact damage amount, temperature rise rate and environmental corrosion index to construct dimensional feature matrix.

5. The contact life prediction method based on big data according to claim 1, characterized in that: The step S4 includes the following sub-steps: Step S401: deploy a lightweight model at the edge to obtain the vibration correction factor and resistance change rate in real time; In step S402 , the cloud receives edge feature data and generates a remaining life probability density function based on Monte Carlo simulation.

6. The contact life prediction method based on big data according to claim 5, characterized in that: The logic of the digital twin verification method in step S5 is: Microcrack extension data, oxide layer thickening data and arc erosion data are collected, weights are set for the microcrack extension data, oxide layer thickening data and arc erosion data respectively, and the reward function is obtained. The reward function is input as reinforcement learning input data into the attention mechanism long short-term memory neural network.

7. The contact life prediction method based on big data according to claim 6, characterized in that: The method further includes step S6, wherein when the salt mist concentration data is greater than a preset salt mist concentration threshold, switching to a preset corrosion-dominated prediction model is performed.

Images