Self-compensating contact resistance on-line monitoring device and method in high-temperature environment

Abstract

The invention relates to a self-compensating contact resistance on-line monitoring device and method in a high-temperature environment, and the device comprises a contact point which is used for forming a contact resistance to be measured; the thermocouple is used for monitoring the temperature in the high-temperature box in real time; the constant current source system is used for applying a constant test current to the contact point and dynamically adjusting an output current value; the signal conditioning circuit is used for filtering high-frequency interference in the voltage signal; the data acquisition card is used for acquiring voltage signals at high speed; and the processor is used for calculating the real value of the compensated contact resistance. The LSTM network based on Kalman filtering and physical constraint has the beneficial effects that in the LSTM network based on Kalman filtering and physical constraint, the KF outputs the denoised base value state in real time through dynamic fusion observation value and model prediction, and clean input is provided for the LSTM. The LSTM models a temperature-resistance nonlinear relation based on a denoising result of the KF, and makes up for a linear hypothesis defect of the KF.

Description

Technical Field

[0001] The present invention relates to the field of on-line monitoring of contact resistance, and more precisely, to an on-line monitoring device and method for contact resistance with self-compensation in a high-temperature environment. Background Art

[0002] Contact resistance is one of the key indicators of electrical connectors, used to characterize the quality of the contact performance of electrical connectors. When monitoring the contact resistance of the contact point in a high-temperature environment, due to the temperature-resistance coupling effect, the resistance change consists of two parts: one part is the reversible change, that is, the resistance change caused by temperature fluctuations, and the other part is the irreversible change, that is, the permanent increase in resistance caused by material oxidation and wear, resulting in the degradation of contact performance. Contact resistance monitoring needs to remove the resistance change caused by temperature fluctuations and retain the contact resistance degradation amount. When the degradation value reaches the failure threshold, it is determined that the contact point fails. Conventional contact resistance monitoring is carried out after the test equipment stops, waiting for the contact point to cool to room temperature, and waiting for the resistance change caused by temperature fluctuations to naturally disappear, and then using an ammeter and a voltmeter to calculate according to Ohm's law. However, this method cannot realize on-line monitoring of contact resistance, cannot capture the instantaneous break event of the contact point, and in addition, the operation is troublesome, increasing the test cost.

[0003] Kalman filtering (KF) is based on a linear state transition and observation model. However, due to thermal expansion hysteresis and material phase change, which will cause non-linear response of resistance, it cannot be accurately modeled by KF. At the same time, the process noise covariance and the observation noise covariance need to be set manually, and it has poor robustness under non-Gaussian noise and time-varying noise. KF also lacks the ability of long-term modeling and cannot capture the aging trend, resulting in unreliable long-term prediction. When directly inputting the noisy original signal into LSTM, high-frequency noise will cause the model to overfit the transient fluctuations, increasing the steady-state prediction error. Pure data-driven LSTM may generate predictions that violate physical laws. Summary of the Invention

[0004] The purpose of the present invention is to propose an on-line monitoring device and method for contact resistance with self-compensation in a high-temperature environment in view of the deficiencies of the prior art.

[0005] In a first aspect, an on-line monitoring device for contact resistance with self-compensation in a high-temperature environment is provided, including:

[0006] A contact point for forming a contact resistance to be measured; the contact point is arranged in a high-temperature box and high-temperature-resistant wires are connected to both ends of the contact point;

[0007] A thermocouple for real-time monitoring of the temperature inside the high-temperature box; the output end of the thermocouple is connected to the processor;

[0008] A constant current source system for applying a constant test current to the contact point and dynamically adjusting the output current value; the constant current source system is connected to the processor;

[0009] A signal conditioning circuit for filtering out high-frequency interference in the voltage signal; the signal conditioning circuit is connected to the contact point;

[0010] A data acquisition card for high-speed acquisition of voltage signals; the data acquisition card is connected to the signal conditioning circuit;

[0011] A processor for calculating the initial value of the contact resistance of the contact point and compensating for the increment of the contact resistance caused by temperature change to obtain the true value of the compensated contact resistance.

[0012] Preferably, the processor includes:

[0013] A calculation module for calculating the initial value of the contact resistance according to the current value of the constant current source system and the voltage value of the data acquisition card;

[0014] A compensation module, based on the Kalman filter and the long short-term memory network, combines the temperature data of the thermocouple to compensate for the resistance change caused by temperature and outputs the true value of the contact resistance.

[0015] Preferably, it further includes: a feedback module; the input end of the feedback module is connected to the output end of the compensation module of the processor, and according to the preset correspondence table between the true value gradient of the contact resistance and the measured current value, outputs a feedback signal of the measured current value corresponding to the current true value gradient of the contact resistance to the input end of the constant current source system.

[0016] Preferably, it further includes: an instantaneous break monitoring module, the input end of the instantaneous break monitoring module is connected to the output end of the compensation module of the processor, and judges whether the true value of the contact resistance exceeds the instantaneous break resistance threshold and the duration exceeds the time threshold according to the instantaneous break resistance threshold and the duration threshold.

[0017] Preferably, the instantaneous break monitoring module is also connected to an alarm.

[0018] In a second aspect, an on-line monitoring method for contact resistance with self-compensation in a high-temperature environment is provided, which is executed by the device according to any one of the first aspects, and includes:

[0019] S1. Apply a constant current to the contact point through the constant current source system and collect the voltage value;

[0020] S2. The calculation module obtains the initial value of the contact resistance according to the current and voltage values;

[0021] S3. The compensation module combines the temperature value of the thermocouple, uses the Kalman filter and the physically constrained LSTM network to compensate for the increment of the contact resistance caused by temperature change, and outputs the true value of the contact resistance;

[0022] S4. The feedback module outputs a feedback signal corresponding to the measured current value of the current true contact resistance gradient to the input end of the constant current source system according to the preset correspondence table between the true contact resistance gradient and the measured current value; the input end of the constant current source system dynamically adjusts the output current value according to the feedback signal;

[0023] S5. The instantaneous break monitoring module determines whether the true contact resistance exceeds the instantaneous break resistance threshold. If the true contact resistance exceeds the instantaneous break resistance threshold and the duration exceeds the time threshold, an alarm is triggered.

[0024] Preferably, S3 includes:

[0025] S301. Remove the random noise of the temperature signal through weighted averaging and retain the low-frequency trend;

[0026] S302. Use the central difference method to calculate the change rates of the denoised temperature estimate and the preliminary contact resistance value;

[0027] S303. Perform Kalman filter state estimation, and separate the base resistance and base temperature from the noise observations through multivariate joint estimation;

[0028] S304. Distinguish whether the current state belongs to the transient stage or the steady state stage through a dynamic gating mechanism and generate a mixed input for the LSTM network;

[0029] S305. Predict the reversible resistance change caused by temperature through a physically constrained LSTM network and output the true contact resistance value.

[0030] Preferably, in S304, the LSTM network includes a short-term memory layer and a long-term memory layer.

[0031] Preferably, in S4, the input end of the constant current source system dynamically adjusts the output current value according to the feedback signal, including: if the current contact resistance increases, feedback to the constant current source system to increase the current output; if the current contact resistance decreases, feedback to the constant current source system to decrease the current output.

[0032] The beneficial effects of the present invention are:

[0033] 1. In the LSTM network based on Kalman filtering and physical constraints of the present invention, KF dynamically fuses the observed values with the model prediction, and outputs the denoised base value state in real time to provide a clean input for the LSTM. Based on the denoising result of KF, the LSTM models the nonlinear relationship between temperature and resistance to make up for the defect of the linear assumption of KF. The input during the transient stage and the steady state stage is adaptively switched through dynamic gating. The oxidation kinetic equation is forcibly injected into the long-term memory unit of the LSTM to ensure that the degradation rate increases exponentially with the increase of temperature. Even when the training data is insufficient, the physical constraints can prevent the model from generating predictions that violate materials science. The noise covariance of KF and the LSTM weights are jointly optimized by gradient descent to adapt to different working conditions.

[0034] 2. The entire measurement process of the present invention does not require stopping the test equipment and special measurement after cooling, saving test time; the compensation module adopted adaptively corrects the contact resistance according to the temperature in the current high-temperature box, improving the accuracy of online monitoring of the contact resistance; the feedback model adopted by the present invention enables the constant current source to adaptively adjust the measurement current according to the fluctuation of the contact resistance; furthermore, the instantaneous break state of the contact resistance is monitored in real time during the measurement process, avoiding the disadvantage of being unable to capture instantaneous break events under static measurement.

[0035] 3. The contact resistance is a key indicator for evaluating the life and reliability of electrical connectors. The static timing measurement method cannot capture the fluctuation response of the contact resistance in real time, ignoring the impact of instantaneous break events on the life and reliability of electrical connectors, and is time-consuming and laborious. The self-compensated online monitoring device and method for contact resistance of the present invention can greatly reduce the test cost, obtain more test data, and improve the efficiency of evaluating the life of electrical connectors. Description of the Drawings

[0036] Figure 1 It is a schematic structural diagram of the self-compensated online monitoring device for contact resistance under high-temperature environment provided by this application;

[0037] Figure 2 It is a flowchart of the self-compensated online monitoring method for contact resistance under high-temperature environment provided by this application;

[0038] Figure 3 It is a flowchart for outputting the true value of the contact resistance;

[0039] Description of the reference numerals: 1 constant current source system, 2 signal conditioning circuit, 3 data acquisition card, 4 contact point, 5 thermocouple, 6 processor, 61 calculation module, 62 compensation module, 7 feedback module, 8 instantaneous break monitoring module, 9 alarm, 10 high-temperature box. Detailed Embodiments

[0040] The present invention will be further described below in conjunction with embodiments. The description of the following embodiments is only for helping to understand the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several modifications can still be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

[0041] Embodiment 1:

[0042] To solve the problems of the prior art, Embodiment 1 of the present application provides an on-line monitoring device for contact resistance with self-compensation under high-temperature environment, including:

[0043] Contact point 4, used to form the contact resistance to be measured; the contact point 4 is arranged in the high-temperature box 10 and high-temperature resistant (temperature resistant above 500 °C) wires are welded at both ends of the contact point 4.

[0044] Thermocouple 5, used to monitor the temperature inside the high-temperature box 10 in real time; the output end of the thermocouple 5 is connected to the processor 6. Exemplarily, the thermocouple 5 is arranged near the contact point 4 in the high-temperature box 10. For example, the distance between the thermocouple 5 and the contact point 4 does not exceed 10 mm.

[0045] Constant current source system 1, used to apply a constant test current to the series loop of contact point 4 - constant current source system 1 - processor 6; the constant current source system 1 is arranged outside the high-temperature box 10 and the output end of the constant current source system 1 is connected to the current application wire of the contact point 4 and the input end of the processor 6;

[0046] Signal conditioning circuit 2, used to filter out high-frequency interference in the voltage signal; the signal conditioning circuit 2 is arranged outside the high-temperature box 10 and is connected to the voltage detection wire of the contact point 4, and the output end is connected to the data acquisition card 3;

[0047] Data acquisition card 3, used to collect voltage signals at high speed; the data acquisition card 3 is arranged after the signal conditioning circuit 2, and the input end is connected to the signal conditioning circuit 2; specifically, the sampling rate of the data acquisition card 3 is at least above 250 kHz, and can capture voltage values at high speed, used to capture contact point instantaneous break events at the 10 μs level.

[0048] Processor 6, used to calculate the preliminary value of the contact resistance of the contact point 4 and compensate for the contact resistance increment caused by temperature change to obtain the true value of the compensated contact resistance. The input end of the processor 6 is connected to the output end of the thermocouple 5, the output end of the constant current source system 1, the current application wire of the contact point 4, and the output end of the data acquisition card 3, where the input end of the processor 6, the output end of the constant current source system 1, and the current application wire of the contact point 4 form a series loop; the output end of the processor 6 is connected to the input end of the constant current source system 1, the alarm 9, the display device and / or the storage device.

[0049] The processor 6 is further configured to obtain the ambient temperature value and the preliminary value of the contact resistance at the contact point 4 through the thermocouple 5, suppress the input noise by Kalman filtering, output the denoised base value state, balance the transient and steady state by dynamic gating, separate the transient effect and the long-term degradation trend through a dual-scale memory structure, and simultaneously embed an oxidation kinetics equation network to ensure that the long-term degradation conforms to physical laws, and output the true value of the contact resistance.

[0050] The processor includes:

[0051] A calculation module 61, configured to calculate the preliminary value of the contact resistance according to the current value of the constant current source system 1 and the voltage value of the data acquisition card 3.

[0052] Specifically, the first input terminal and the second input terminal of the calculation module 61 form a series circuit with the output terminal of the constant current source system 1 and the current application wire of the contact point 4, and the third input terminal is connected to the output terminal of the data acquisition card 3, and is configured to calculate the preliminary value of the contact resistance in real time and online according to the current value of the constant current source system 1 and the voltage value of the data acquisition card 3.

[0053] A compensation module 62, based on Kalman filtering and a long short-term memory network, combines the temperature data of the thermocouple 5 to compensate for the resistance change caused by temperature, and outputs the true value of the contact resistance.

[0054] Specifically, the first input terminal of the compensation module 62 is connected to the output terminal of the calculation module 61, and the second input terminal is connected to the output terminal of the thermocouple 5. According to the preliminary value of the contact resistance of the calculation module 61 and the ambient temperature value of the contact point 4 obtained by the thermocouple 5, the contact resistance increment caused by temperature change is compensated through an LSTM network based on Kalman filtering and physical constraints, and the true value of the compensated contact resistance is obtained. The output terminal is externally connected to a display device and / or a storage device for displaying and / or storing the true value of the contact resistance.

[0055] Embodiment 2:

[0056] On the basis of Embodiment 1, Embodiment 2 of the present application provides a more specific on-line monitoring device for contact resistance with self-compensation in a high-temperature environment, including: a contact point 4, a thermocouple 5, a constant current source system 1, a data acquisition card 3, and a processor 6.

[0057] Wherein, at least one thermocouple 5 is provided, and the data acquisition card 3 does not directly collect the voltage value of the contact point 4, and the high-frequency interference is filtered by the signal conditioning circuit 2 and then input into the data acquisition card 3. In specific implementation, the contact point 4 and the thermocouple 5 can be placed in a constant temperature test chamber, a thermal cycle test chamber, and a temperature-vibration comprehensive test chamber. The thermocouple 5 can be a commonly used thermocouple on the market.

[0058] The contact point 4 is in series with the constant current source system 1 and the calculation module 61 in the processor 6 through a high-temperature resistant wire, and the constant current source system 1 supplies the measurement current. The contact point 4 is connected to the signal conditioning circuit 2 through a high-temperature resistant wire. After the voltage measured at the contact point 4 end of the signal conditioning circuit 2 is high-frequency filtered, it is transmitted to the data acquisition card 3. If there are multiple contact points, they can be connected in series and supplied with current by the constant current source system 1 uniformly. The data acquisition card 3 can be a multi-channel 16-bit data acquisition card.

[0059] The calculation module 61 calculates the initial value of the contact resistance based on the voltage value collected by the acquisition card 2 and the current value supplied by the constant current source system 1. However, the initial value of the contact resistance contains interference caused by high temperature.

[0060] The compensation module 62 compensates for the increment of the contact resistance caused by temperature change based on the Kalman filter and the LSTM network with physical constraints, and obtains the true value of the compensated contact resistance.

[0061] In specific implementation, the processor uses a multi-core processor, such as the ARM Cortex-A series.

[0062] The processor 6 is connected to a display device and / or a storage device, and displays and / or stores the true value of the contact resistance in real time.

[0063] In specific implementation, different display devices or storage devices can be selected according to actual needs, such as a liquid crystal display or an SD memory card.

[0064] In addition, it further includes: a feedback module 7; the input end of the feedback module 7 is connected to the output end of the compensation module 62 of the processor 6. According to the preset correspondence table between the true value gradient of the contact resistance and the measured current value, it outputs a feedback signal corresponding to the measured current value of the current true value gradient of the contact resistance to the input end of the constant current source system 1. The input end of the constant current source system 1 is connected to the feedback module 7 and receives the feedback signal to dynamically adjust the output current value. If the current contact resistance increases, it feeds back to the constant current source system 1 to increase the current output. If the current contact resistance decreases, it feeds back to the constant current source system 1 to decrease the current output. For example, when the contact point is in a high-resistance state, the current is increased to 100 mA, and when it is in a low-resistance state, it is decreased to 10 mA.

[0065] It further includes: an instantaneous break monitoring module 8. The input end of the instantaneous break monitoring module 8 is connected to the output end of the compensation module 62 of the processor 6. According to the instantaneous break resistance threshold and the duration threshold, it judges whether the true value of the contact resistance exceeds the instantaneous break resistance threshold and the duration exceeds the time threshold. The instantaneous break monitoring module 8 is also connected to an alarm 9. If the true value of the contact resistance exceeds the instantaneous break resistance threshold and the duration exceeds the time threshold, an alarm for the instantaneous break event of the contact point is given.

[0066] In specific implementation, the instantaneous break resistance threshold and the duration threshold can be set according to actual engineering requirements. For example, the contact resistance , and the duration are the thresholds for instantaneous break events.

[0067] During the entire measurement process, only the current application wire and the voltage detection wire of the contact point 4 and the detection wire of the thermocouple 5 need to be led out through the 10-wire hole of the high-temperature box and connected to the on-line monitoring device, so as to realize the on-line test of the contact resistance in the high-temperature or temperature cycling or temperature vibration state. The measurement result can filter out the influence of interference signals and temperature drift, improve the accuracy of the measurement result, and can also realize the alarm of the instantaneous break event of the contact point 4, reduce the monitoring cost, and the monitoring process is also very simple.

[0068] It should be noted that the same or similar parts in this embodiment and Embodiment 1 can be referred to each other and will not be elaborated in this application.

[0069] Embodiment 3:

[0070] Based on Embodiment 2, Embodiment 3 of the present application provides an on-line monitoring method for contact resistance with self-compensation in a high-temperature environment, as Figure 2 shown, including:

[0071] S1. Apply a constant current to the contact point through a constant current source system and collect the voltage value.

[0072] S2. The calculation module obtains a preliminary value of the contact resistance according to the current and voltage values.

[0073] S3. The compensation module combines the temperature value of the thermocouple and uses the Kalman filter and the LSTM network with physical constraints to compensate for the increment of the contact resistance caused by temperature changes and outputs the true value of the contact resistance.

[0074] As Figure 3 shown, S3 includes:

[0075] S301. Remove the random noise of the temperature signal through weighted averaging and retain the low-frequency trend.

[0076] Specifically, the ambient temperature value of the contact point 4 obtained by the thermocouple 5 contains ambient noise and electromagnetic interference, and the preliminary value of the contact resistance obtained by the calculation module 61 contains the reversible resistance change caused by temperature and the irreversible change caused by degradation, which is expressed as:

[0077]

[0078] Among them, is the resistance change caused only by irreversible degradation and is a long-term trend, It is a reversible change caused by temperature fluctuations and is a short-term fluctuation. The random noise of the temperature signal is removed by weighted averaging, and the low-frequency trend is retained, expressed as:

[0079]

[0080] In the formula, is the sliding window weight coefficient, satisfying , and is used to suppress high-frequency noise; n is the window radius, which controls the smoothing intensity.

[0081] S302. Calculate the denoised temperature estimate value and the initial value of the contact resistance The rate of change is captured to characterize the dynamic change of the signal and provide a basis for transient detection, expressed as:

[0082]

[0083] In the formula, is the time step, which is used for central difference to calculate the first-order derivative.

[0084] S303. Perform Kalman filter state estimation. Through multi-variable joint estimation, the base resistance and base temperature are separated from the noisy observations. At the same time, the resistance temperature coefficient is used to suppress noise and output the preliminary denoised result.

[0085] Specifically, define the state vector:

[0086]

[0087] In the formula, is the base resistance, which represents the true resistance after removing the temperature effect and only reflects the degradation state; is the base temperature, which represents the stable temperature estimate after removing noise; , is the rate of change of the base state.

[0088] Define the state transition model:

[0089]

[0090] In the formula, is the state transition matrix, which models the dynamic evolution of the state variables. Preferably, a first-order motion model is adopted; is the process noise, and the covariance matrix Q represents the uncertainty of the model prediction.

[0091] Define the observation model:

[0092]

[0093] In the formula, is the observation matrix, which defines the linear mapping relationship between the state variable and the observation signal; is the observation noise, and the covariance matrix R represents the temperature value measurement error.

[0094] Define the update step:

[0095]

[0096] In the formula, is the Kalman gain, which balances the weights of the predicted value and the observed value and is dynamically calculated through the covariance matrix

[0097] S304. Distinguish whether the current state belongs to the transient stage or the steady state stage through the dynamic gating mechanism, and generate the mixed input of the LSTM network.

[0098] In S304, calculate the gating coefficient, which is expressed as:

[0099]

[0100] In the formula, is the Sigmoid function, and the output range is (0, 1); is the slope coefficient, which controls the steepness of the gating switch; is the threshold value. When the sum of the signal change rates exceeds , it is determined to be the transient stage.

[0101] Generate the mixed input of the LSTM network, which is expressed as:

[0102]

[0103] When it is currently in the transient stage , the LSTM network preferentially uses the KF denoising result to suppress noise interference; when it is currently in the dynamic stage , the LSTM network uses the signal that retains the original details to avoid the phase delay introduced by filtering.

[0104] In S304, the LSTM network includes a short-term memory layer and a long-term memory layer. Among them, the short-term memory layer captures the second-level fluctuations, such as the transient resistance change caused by the temperature mutation, which is expressed as:

[0105]

[0106] The long-term memory layer is:

[0107] ,

[0108] In the formula, ​The forgetting gate is physically constrained and modulated by temperature; The input gate controls the proportion of new information written.

[0109] Physical constraint The specific design is as follows:

[0110]

[0111] In the formula, is the activation energy of the oxidation reaction; R is the gas constant; is the base temperature for KF estimation. When the temperature rises, the forgetting gate coefficient approaches 1, retaining more historical memories and simulating the oxidation accumulation effect.

[0112] Calculate the regularization loss term to force the long-term memory update rate to conform to the Arrhenius equation and ensure the physical rationality of degradation prediction, expressed as:

[0113]

[0114] In the formula, k is the oxidation reaction rate constant.

[0115] S305. Predict the reversible resistance change caused by temperature through the physically constrained LSTM network and output the true value of the contact resistance.

[0116] Specifically, the final output result in S305 includes the reversible resistance change caused by temperature predicted by the LSTM , expressed as:

[0117]

[0118] In the formula, is the weight matrix, which fuses short-term and long-term temperature features and quantifies the dynamic contributions of both to the resistance change; is the bias term, which provides a temperature-independent resistance reference offset for correcting the baseline value of the model output; is the long-term feature, representing the historical accumulation effect or the slow-changing trend.

[0119] Deduct the temperature effect from the original resistance to obtain the base resistance that only reflects degradation This is the true value of the contact resistance, expressed as:

[0120]

[0121] S4. The feedback module outputs a feedback signal of the measured current value corresponding to the current true value gradient of the contact resistance to the input end of the constant current source system according to the preset correspondence table between the true value gradient of the contact resistance and the measured current value; the input end of the constant current source system dynamically adjusts the output current value according to the feedback signal.

[0122] In S4, the input end of the constant current source system dynamically adjusts the output current value according to the feedback signal, including: if the current contact resistance increases, it feeds back to the constant current source system to increase the current output; if the current contact resistance decreases, it feeds back to the constant current source system to decrease the current output.

[0123] S5. The instantaneous break monitoring module determines whether the true value of the contact resistance exceeds the instantaneous break resistance threshold. If the true value of the contact resistance exceeds the instantaneous break resistance threshold and the duration exceeds the time threshold, an alarm is triggered.

[0124] It should be noted that the method provided in this embodiment is the method corresponding to the device provided in Embodiment 2. Therefore, for the parts that are the same or similar in this embodiment and Embodiment 2, reference can be made to each other and will not be elaborated in this application.

[0125] Embodiment 4:

[0126] Based on Embodiment 3, Embodiment 4 of this application provides a more specific on-line monitoring method for contact resistance with self-compensation under high-temperature environment, including:

[0127] 1. Input signal and preprocessing (t = 10s)

[0128] 1) The original signal is as follows

[0129] , with ±2°C noise; , with ±5mΩ measurement noise.

[0130] 2) Perform preprocessing on through moving average filtering, and select the window weight as to obtain the temperature estimated value after removing high-frequency noise

[0131]

[0132] 3) Take , and use the central difference method to calculate the change rate

[0133]

[0134] 2. Kalman filter (KF) state estimation

[0135] 1) KF initialization (t = 9s state)

[0136]

[0137] The diagonal elements of the covariance matrix represent the uncertainty of the state variables, and The variance is 0.1, the variance of the rate of change is 0.01, and the covariance matrix is expressed as:

[0138]

[0139] 2) Prediction step (t = 10s). Assume that the base value resistance and temperature change linearly with time, like a first-order motion model. The state transition equation is:

[0140]

[0141] where the state transition matrix F is:

[0142]

[0143] Update the prediction covariance:

[0144]

[0145] The process noise takes , indicating the uncertainty of the model prediction.

[0146] 3) Update step (t = 10s). First, obtain the observation vector:

[0147]

[0148] Then calculate the observation residual:

[0149]

[0150] The second row of the observation matrix H , where α = 0.004 is the temperature coefficient of resistance. The observation matrix H is

[0151]

[0152] 3. Dynamic gating

[0153] Calculate the gating coefficient:

[0154]

[0155] , , when , the gating coefficient approaches 1. The current is the transient stage, and the KF result is preferably used to suppress noise.

[0156] Generate the mixed input , avoiding the interference of the original noise to the LSTM network.

[0157]

[0158] 4. LSTM Network and Physical Constraints

[0159] The short - term memory layer models the second - level fluctuations and inputs the mixed features and outputs a 32 - dimensional hidden state to capture the transient response of resistance caused by temperature mutations.

[0160] The long - term memory layer models the hourly - level aging. First, the forget gate is modulated as:

[0161]

[0162] At this time, the temperature , high temperature accelerates oxidation, and the forget gate retains more historical memories to simulate the accumulation of degradation.

[0163] Update the cell state , is the candidate value of the new information at the current moment, expressed as:

[0164]

[0165] Verify the physical constraints, expressed as:

[0166]

[0167] When the update rate of the long - term memory deviates from the theoretical value, force the LSTM to adjust the weights through the gradient so that the update rate of the long - term memory approaches the theoretical oxidation rate.

[0168] Decouple the temperature effect to obtain the change in resistance caused by the temperature predicted by the LSTM, is the weight matrix, is the bias term, which is obtained by fitting the training data:

[0169]

[0170] Finally, calculate the base - value resistance, that is, the true value of the contact resistance, expressed as:

[0171]

[0172] Compared with the of KF, the LSTM is further corrected through non - linear modeling.

[0173] All parameters (Q, R of KF, LSTM weights, gating coefficients) are jointly optimized through end - to - end training. The above embodiments are just one of the cases.

[0174] It should be noted that the parts that are the same or similar to those in Embodiment 3 in this embodiment can be referred to each other and will not be elaborated in this application.

Claims

1. An on-line monitoring device for contact resistance with self-compensation under high temperature environment, characterized in that, Comprising: Contact points for forming the contact resistance to be measured; The contact points are arranged inside the high-temperature chamber and high-temperature resistant wires are connected to both ends of the contact points; Thermocouples for real-time monitoring of the temperature inside the high-temperature chamber; the output ends of the thermocouples are connected to the processor; A constant current source system for applying a constant test current to the contact points and dynamically adjusting the output current value; The constant current source system is connected to the processor; A signal conditioning circuit for filtering high-frequency interference in the voltage signal; The signal conditioning circuit is connected to the contact points; A data acquisition card for high-speed acquisition of voltage signals; the data acquisition card is connected to the signal conditioning circuit; A processor for calculating the preliminary value of the contact resistance of the contact points and compensating for the increment of the contact resistance caused by temperature changes to obtain the true value of the compensated contact resistance.

2. The on-line monitoring device for contact resistance with self-compensation under high temperature environment according to claim 1, wherein, The processor includes: A calculation module for calculating the preliminary value of the contact resistance according to the current value of the constant current source system and the voltage value of the data acquisition card; A compensation module, based on Kalman filtering and long short-term memory network, combines the temperature data of the thermocouple to compensate for the resistance change caused by temperature and outputs the true value of the contact resistance.

3. The on-line monitoring device for contact resistance with self-compensation under high temperature environment according to claim 2, wherein It further includes: A feedback module; The input end of the feedback module is connected to the output end of the compensation module of the processor. According to the preset correspondence table between the true value gradient of the contact resistance and the measured current value, it outputs a feedback signal of the measured current value corresponding to the current true value gradient of the contact resistance to the input end of the constant current source system.

4. The on-line monitoring device for contact resistance with self-compensation under high temperature environment according to claim 3, characterized in that It further includes: An instantaneous break monitoring module, the input end of the instantaneous break monitoring module is connected to the output end of the compensation module of the processor. According to the instantaneous break resistance threshold and the duration threshold, it judges whether the true value of the contact resistance exceeds the instantaneous break resistance threshold and the duration exceeds the time threshold.

5. The on-line monitoring device for contact resistance with self-compensation under high temperature environment according to claim 4, characterized in that, The instantaneous break monitoring module is also connected to an alarm.

6. An on-line monitoring method for contact resistance with self-compensation under high temperature environment, characterized in that, Executed by the device according to any one of claims 1 to 5, including: S1. Apply a constant current to the contact points through the constant current source system and collect the voltage value; S2. The calculation module obtains the preliminary value of the contact resistance according to the current and voltage values; S3. The compensation module combines the temperature value of the thermocouple, uses Kalman filtering and the LSTM network with physical constraints to compensate for the increment of the contact resistance caused by temperature changes and outputs the true value of the contact resistance; S4. The feedback module outputs a feedback signal of the measured current value corresponding to the current true value gradient of the contact resistance to the input end of the constant current source system according to the preset correspondence table between the true value gradient of the contact resistance and the measured current value; the input end of the constant current source system dynamically adjusts the output current value according to the feedback signal; S5. The instantaneous break monitoring module judges whether the true value of the contact resistance exceeds the instantaneous break resistance threshold. If the true value of the contact resistance exceeds the instantaneous break resistance threshold and the duration exceeds the time threshold, an alarm is triggered.

7. The on-line monitoring method for contact resistance with self-compensation under high temperature environment according to claim 6, wherein S3 Including: S301. Remove the random noise of the temperature signal by weighted average and retain the low-frequency trend; S302. Use the central difference method to calculate the change rates of the denoised temperature estimate value and the preliminary value of the contact resistance; S303. Perform Kalman filter state estimation, and through multi-variable joint estimation, separate the base resistance and base temperature from the noise observation; S304. Distinguish whether the current state belongs to the transient stage or the steady state stage through a dynamic gating mechanism and generate a mixed input for the LSTM network; S305. Predict the reversible resistance change caused by temperature through a physically constrained LSTM network and output the true value of the contact resistance.

8. The on-line monitoring method for contact resistance with self-compensation under high temperature environment according to claim 7, characterized in that, In S304, the LSTM network includes a short-term memory layer and a long-term memory layer.

9. The on-line monitoring method for contact resistance with self-compensation under high temperature environment according to claim 8, characterized in that In S4, the input end of the constant current source system dynamically adjusts the output current value according to the feedback signal, including: if the current contact resistance increases, it feeds back to the constant current source system to increase the current output; if the current contact resistance decreases, it feeds back to the constant current source system to decrease the current output.

Images