Portable relay calibration device and life prediction method using the same

Abstract

The present invention discloses a portable relay calibration device and a life prediction method using the device, relating to the technical field of detection. The device comprises: a battery pack, an inverter, an industrial computer, a multimeter, a switching power supply, a switch, a switching device, a relay interface and a battery charger. The relay is connected to the calibration device. Through the instructions of the industrial computer to the multimeter and the switching device, the device can realize the measurement of the relay coil resistance, contact resistance, action voltage and action time. The life prediction method is implemented based on the calibration device. The life prediction method realizes the prediction of the remaining life of the relay by establishing a complex neural network model, combining data such as the number of relay actions, action time and contact resistance to construct a relationship through curve fitting.

Description

Technical Field

[0001] The present invention relates to the technical field of relay detection, and in particular to a portable relay calibration device and a life prediction method using the device. Background Art

[0002] The existing portable relay testing device consists of a battery switching power supply, a transfer switch, a DC voltmeter, an ohmmeter, an industrial computer, and host computer software. It can detect the operating voltage and release voltage, coil resistance, and contact resistance of the relay.

[0003] However, existing portable relay test devices require an external power supply and cannot be used in situations where there is no power supply. They cannot automatically generate test reports and statistical analysis data and can only determine whether the relay is qualified, but cannot determine the health status of the relay.

[0004] In addition, predicting the life of an existing relay requires collecting the relay's operating waveform, which means that an oscilloscope needs to be configured to extract the waveform's characteristic information through a complex algorithm. Summary of the Invention

[0005] The purpose of the present invention is to realize the life prediction of a portable relay.

[0006] The technical solution of the present invention is: providing a portable relay calibration device, the device comprising: a battery pack, an inverter, an industrial computer, a multimeter, a switching power supply, a switch, a switching device, a relay interface and a battery charger;

[0007] The battery pack converts DC power into AC power through an inverter and inputs it into the power supply port of the industrial computer, the power supply port of the multimeter, the input port of the switching power supply, and the power supply port of the switch. The battery pack powers the industrial computer, the multimeter, and the switch, and provides trigger current for the switching power supply.

[0008] After the input port of the switching power supply receives the alternating current as the trigger current, the switching power supply provides direct current to the switching device through the output port; the measurement port of the multimeter is connected to the switching device; the switching device is connected to the terminals of the relay interface, and the relay interface is connected to the relay; the switching device is used to bridge the measurement port of the multimeter and the terminals of the relay interface;

[0009] The industrial computer is connected to the multimeter, switching power supply and switching device through the switch to realize network communication.

[0010] In any of the above technical solutions, further, the multimeter has three measurement ports: a first measurement port, a second measurement port and a third measurement port, and the first measurement port, the second measurement port and the third measurement port are connected to different input ports of the switching device; the relay interface includes three terminals: a coil terminal, a normally closed contact terminal and a normally open contact terminal, and the coil terminal, the normally closed contact terminal and the normally open contact terminal are respectively connected to different output ports of the switching device; in the switching device, the multimeter connects the first measurement port, the second measurement port and the third measurement port to the coil terminal, the normally closed contact terminal and the normally open contact terminal of the relay interface according to preset rules according to the specific instructions received.

[0011] In any of the above technical solutions, further, the relay verification device includes five relay interfaces, the same terminal of each relay interface is connected to the same port of the switching device, and each interface performs testing independently.

[0012] In any of the above technical solutions, further, a method for using the device includes:

[0013] The relay is connected to the relay interface of the switching device. The industrial computer sends a command to the multimeter. The multimeter switches the channels of the first measurement port, the second measurement port, and the third measurement port to the resistance measurement position. The industrial computer also sends a command to the switching device. The switching device connects the first measurement port to the coil terminal and the second measurement port to the normally closed contact terminal. The channel of the multimeter where the first measurement port is located measures the coil resistance of the relay, and the channel where the second measurement port is located measures the normally closed contact resistance of the relay. The multimeter sends the measurement data to the industrial computer.

[0014] After receiving the measurement data, the industrial computer sends a command to the multimeter to switch the channel where the first measurement port is located to the voltage measurement position. The switching device introduces the DC power output by the switching power supply into the first measurement port. When the channel where the second measurement port of the multimeter is located detects the actuation of the relay contact, that is, the resistance value of the normally closed contact becomes infinite, the voltage measured at the channel where the first measurement port of the multimeter is located is sent to the industrial computer as the actuation voltage of the relay.

[0015] The switching power supply continues to gradually increase its output voltage. When the output voltage of the switching power supply reaches the rated voltage of the relay and stops increasing, the industrial computer sends a command to the switching device to connect the third measuring port to the normally open contact terminal. The channel corresponding to the third measuring port of the multimeter measures the normally open contact resistance and sends the result to the industrial computer. The multimeter calculates the relay actuation time based on the time from the third measuring port being connected to the normally open contact terminal until the measured resistance stabilizes.

[0016] The switching power supply gradually reduces its output voltage. During this process, the channel where the third measurement port of the multimeter is located detects the action of the relay contact. That is, when the resistance of the normally open contact becomes infinite, the voltage measured by the channel where the first measurement port of the multimeter is located at this time is sent to the industrial computer as the release voltage of the relay.

[0017] After completing the above measurements, the host computer displays the test verification results through a standard report and enters the test results into the relay database for subsequent statistical analysis.

[0018] In any of the above technical solutions, further, when the relay verification device is charging, the battery pack is connected to the charging port through the battery charger to replenish power.

[0019] A method for predicting the life of a portable relay using the portable relay calibration device is also provided, the method comprising:

[0020] A lifespan prediction model was established, which consists of an input layer, a TCN layer, an LSTM layer, an attention layer, and an output layer. The TCN layer uses TCN as a feature extractor. TCN treats the sequence data input from the input layer as a one-dimensional signal, performs convolution operations through multiple layers of convolution kernels, and finally inputs the output results into the LSTM layer.

[0021] The LSTM layer consists of three layers of long short-term memory networks. Each layer contains several time steps. Each time step contains an input gate, a forget gate, and an output gate. The input gate determines how the data of the current time step affects the state of the memory unit; the forget gate determines how the current time step forgets the information of the previous time step; and the output gate determines how the state of the memory unit of the current time step affects the output feature.

[0022] The first LSTM layer receives the output from the TCN layer, captures the long-term dependencies in the input data, and extracts preliminary temporal features. The processed features are passed to the second and third LSTM layers for further extraction and finally to the attention layer.

[0023] The output of the LSTM layer is fed into the attention layer, which includes the attention mechanism (ECA) and a fully connected layer. The ECA module has a relatively simple structure, consisting of three main steps: global average pooling, one-dimensional convolution, and channel weighting. Global average pooling calculates the global average of each channel to obtain a vector representing the importance of the channel. The one-dimensional convolution operation processes this vector and learns the relationship between channels. The fully connected layer connects the output of the attention mechanism and weights the feature map of each channel using the corresponding attention weight. The final result is used as the output of the lifespan prediction model.

[0024] By repeatedly using the relay, the relationship between the number of relay actuations and its lifespan was tested. The relay's actuation time and contact resistance increase with the number of actuations. A brand new relay was connected to the portable relay calibration device described above. The actuation time and contact resistance were recorded every 100 actuations until the relay failed. After the relay failed, the actuation count was recorded. The same procedure was repeated for several identical brand new relays, with all actuation time and contact resistance data recorded and sorted by the number of actuations. All data was normalized.

[0025] The above normalized data is used as a training set to input the life prediction model for training. After the training, the relationship between the number of operations, operation time and contact resistance is obtained by fitting the curve according to the output of the life prediction neural network. After testing the relay, the operation time and contact resistance are substituted into the relationship to calculate the theoretical number of operations, thereby predicting the remaining life of the relay.

[0026] In any of the above technical solutions, further, the expression of the convolution operation of the TCN layer is: ,in For input, Indicates the The parallel layer The output of the convolution kernel, Indicates the size , the expansion rate is The convolution kernel, Represents the causal dilated convolution operation.

[0027] In any of the above technical solutions, further, the life prediction model uses Adam as the optimizer, the learning rate is set to 0.0015, and the As the loss function, represents the number of training samples, represents the true value of the sample, Represents the output value of the network.

[0028] The beneficial effects of the present invention are:

[0029] The technical solution of the present invention provides a portable relay calibration device and a supporting life prediction model, thereby realizing portable relay life prediction; the commonly used life prediction model only uses LSTM for prediction. Although LSTM performs well in processing long time series data, it has high computational complexity, difficulty in parallel computing and long-distance dependency problems; the present invention introduces a temporal convolutional network (TCN) and an efficient channel attention mechanism (ECA) to effectively overcome these shortcomings: TCN expands the receptive field through parallel computing and dilated convolution, improving computational efficiency and stability; ECA enhances feature expression capabilities through a lightweight attention mechanism, thereby improving the overall performance and efficiency of the model, and controlling the final prediction error within 5%. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The advantages of the above and additional aspects of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which:

[0031] Figure 1 is a structural schematic diagram of a portable relay calibration device according to an embodiment of the present invention;

[0032] Figure 2 2. It is a schematic structural diagram of a charger for a portable relay calibration device according to an embodiment of the present invention;

[0033] Figure 3 The figure is a schematic diagram of the life prediction model structure adopted by the life prediction method of the portable relay calibration device according to one embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features therein can be combined with each other without conflict.

[0035] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0036] like Figure 1 As shown, this embodiment provides a portable relay calibration device, which includes: a battery pack, an inverter, an industrial computer, a multimeter, a switching power supply, a switch, a switching device, a relay interface and a battery charger.

[0037] The battery pack converts DC power into AC power through an inverter and inputs it into the power supply port of the industrial computer, the power supply port of the multimeter, the input port of the switching power supply, and the power supply port of the switch. The battery pack powers the industrial computer, the multimeter, and the switch, and provides trigger current for the switching power supply.

[0038] The output scale of the inverter used in this embodiment is 220V AC.

[0039] After the input port of the switching power supply receives alternating current as a trigger current, the switching power supply provides direct current to the switching device through the output port; the measuring port of the multimeter is connected to the switching device; the switching device is connected to the terminals of the relay interface, and the relay interface is connected to the relay; the switching device is used to bridge the measuring port of the multimeter and the terminals of the relay interface.

[0040] The industrial computer is connected to the multimeter, switching power supply and switching device through the switch to realize network communication.

[0041] The multimeter has three measurement ports: a first measurement port, a second measurement port, and a third measurement port, which are connected to different input ports of a switching device; the relay interface includes three terminals: a coil terminal, a normally closed contact terminal, and a normally open contact terminal, which are respectively connected to different output ports of the switching device; within the switching device, the multimeter connects the first measurement port, the second measurement port, and the third measurement port to the coil terminal, the normally closed contact terminal, and the normally open contact terminal of the relay interface according to specific rules based on the specific instructions received.

[0042] When it is necessary to calibrate the relay connected to the relay interface, the industrial computer sends a command to the multimeter, and the multimeter switches the channels where the first measurement port, the second measurement port, and the third measurement port are located to the resistance measurement position. The industrial computer also sends a command to the switching device, and the switching device connects the first measurement port to the coil terminal and the second measurement port to the normally closed contact terminal. The channel where the first measurement port of the multimeter is located measures the coil resistance of the relay, and the channel where the second measurement port is located measures the normally closed contact resistance of the relay. The multimeter sends the measurement data to the industrial computer.

[0043] After receiving the measurement data, the industrial computer sends a command to the multimeter to switch the channel where the first measurement port is located to the voltage measurement position. The switching device introduces the DC power output by the switching power supply into the first measurement port. When the channel where the second measurement port of the multimeter is located detects the action of the relay contact (manifested by the resistance of the normally closed contact becoming infinite), the voltage measured by the channel where the first measurement port of the multimeter is located at this time is sent to the industrial computer as the action voltage of the relay.

[0044] The switching power supply continues to gradually increase its output voltage. When the output voltage of the switching power supply reaches the rated voltage of the relay, it no longer increases. At this time, the industrial computer sends an instruction to the switching device to connect the third measuring port to the normally open contact terminal. The channel where the third measuring port of the multimeter is located measures the normally open contact resistance and sends it to the industrial computer. The multimeter obtains the relay operation time based on the time from the third measuring port to the normally open contact terminal being connected until the measured resistance stabilizes.

[0045] The switching power supply gradually reduces its output voltage. During this process, the channel where the third measurement port of the multimeter is located detects the action of the relay contact (manifested as the resistance of the normally open contact becomes infinite). The voltage measured at the channel where the first measurement port of the multimeter is located is sent to the industrial computer as the release voltage of the relay.

[0046] After completing the above measurements, the host computer displays the test verification results through a standard report and enters the test results into the relay database for subsequent statistical analysis.

[0047] In this embodiment, the relay verification device includes five relay interfaces, and the same terminal of each relay interface is connected to the same port of the switching device, and each interface can perform testing work independently.

[0048] like Figure 2 As shown, when the relay verification device is charged, the battery pack is connected to the charging port through the battery charger to replenish power.

[0049] Another embodiment of the present invention provides a relay life prediction method used in the portable relay verification device, the method comprising:

[0050] like Figure 3 As shown in the figure, a life prediction model is established, which includes an input layer, a TCN layer, an LSTM layer, an attention layer, and an output layer. The TCN layer uses TCN as a feature extractor to capture key features in the time domain (the order of the number of relay operations). TCN regards the sequence data input from the input layer as a one-dimensional signal, performs convolution operations through multiple layers of convolution kernels, and finally inputs the output results into the LSTM layer.

[0051] The expression of the above convolution operation is: ,in For input, Indicates the The parallel layer The output of the convolution kernel, Indicates the size , the expansion rate is The convolution kernel, Represents the causal dilated convolution operation.

[0052] The LSTM layer consists of three layers of long short-term memory networks. Figure 3 The spheres in the LSTM layer represent time steps. Each time step contains an input gate, a forget gate, and an output gate. The input gate determines how the data in the current time step affects the state of the memory cell; the forget gate determines how the current time step forgets the information from the previous time step; and the output gate determines how the state of the memory cell in the current time step affects the output features. The first LSTM layer receives the output from the TCN layer, captures the long-term dependencies in the input data, and extracts preliminary temporal features. These processed features are then passed to the second and third LSTM layers for further extraction, and finally to the attention layer.

[0053] The output of the LSTM layer is fed into the attention layer, which consists of an attention mechanism (ECA) and a fully connected layer. The ECA module has a relatively simple structure, consisting of three main steps: global average pooling, one-dimensional convolution, and channel weighting. Global average pooling calculates the global average of each channel to obtain a vector representing channel importance. The one-dimensional convolution operation processes this vector and learns the relationship between channels. The fully connected layer concatenates the output of the attention mechanism and weights the feature maps of each channel using the corresponding attention weights. This significantly improves model performance without introducing excessive computational overhead, and the final result serves as the output of the lifespan prediction model.

[0054] The above life prediction model uses Adam as the optimizer, the learning rate is set to 0.0015, and As the loss function, represents the number of training samples, represents the true value of the sample, Represents the output value of the network.

[0055] The relationship between the number of relay actuations and lifespan was tested by repeated use of the relay. The relay's actuation time and contact resistance increase with the number of actuations. A brand new relay was connected to the portable relay calibration device described above. The actuation time and contact resistance were recorded every 100 actuations until the relay failed. After failure, the actuation count was recorded. The same procedure was repeated for several identical brand new relays, with all actuation times and contact resistance data recorded and sorted by actuation count. All data was normalized.

[0056] The above normalized data is used as a training set to input the life prediction model for training. After the training, the relationship between the number of operations, operation time and contact resistance is obtained by fitting the curve according to the output of the life prediction neural network. After testing the relay, the operation time and contact resistance are substituted into the relationship to calculate the theoretical number of operations, thereby predicting the remaining life of the relay.

[0057] In summary, the present invention provides a portable relay calibration device, which includes: a battery pack, an inverter, an industrial computer, a multimeter, a switching power supply, a switch, a switching device, a relay interface and a battery charger.

[0058] The battery pack converts DC power into AC power through an inverter and inputs it into the power supply port of the industrial computer, the power supply port of the multimeter, the input port of the switching power supply, and the power supply port of the switch. The battery pack powers the industrial computer, the multimeter, and the switch, and provides trigger current for the switching power supply.

[0059] After the input port of the switching power supply receives alternating current as a trigger current, the switching power supply provides direct current to the switching device through the output port; the measuring port of the multimeter is connected to the switching device; the switching device is connected to the terminals of the relay interface, and the relay interface is connected to the relay; the switching device is used to bridge the measuring port of the multimeter and the terminals of the relay interface.

[0060] The industrial computer is connected to the multimeter, switching power supply and switching device through the switch to realize network communication.

[0061] The present invention also provides a life prediction method using the portable relay calibration device, the method comprising:

[0062] A lifespan prediction model is established, which includes an input layer, a TCN layer, an LSTM layer, an attention layer, and an output layer. The TCN layer uses TCN as a feature extractor. TCN regards the sequence data input from the input layer as a one-dimensional signal, performs convolution operations through multiple layers of convolution kernels, and finally inputs the output results into the LSTM layer.

[0063] The LSTM layer consists of three layers of long short-term memory networks. Each layer contains several time steps. Each time step contains an input gate, a forget gate, and an output gate. The input gate determines how the data of the current time step affects the state of the memory unit; the forget gate determines how the current time step forgets the information of the previous time step; and the output gate determines how the state of the memory unit in the current time step affects the output features.

[0064] The first LSTM layer receives the output from the TCN layer, captures the long-term dependencies in the input data, and extracts preliminary temporal features. The processed features are passed to the second and third LSTM layers for further extraction and finally to the attention layer.

[0065] The output of the LSTM layer is fed into the attention layer, which consists of an attention mechanism (ECA) and a fully connected layer. The ECA module has a relatively simple structure, consisting of three main steps: global average pooling, one-dimensional convolution, and channel weighting. Global average pooling calculates the global average of each channel to obtain a vector representing channel importance. The one-dimensional convolution operation processes this vector and learns the relationship between channels. The fully connected layer concatenates the output of the attention mechanism and weights the feature maps of each channel using the corresponding attention weights. The final result is used as the output of the lifespan prediction model.

[0066] The relationship between the number of relay actuations and lifespan was tested by repeated use of the relay. The relay's actuation time and contact resistance increase with the number of actuations. A brand new relay was connected to the portable relay calibration device described above. The actuation time and contact resistance were recorded every 100 actuations until the relay failed. After failure, the actuation count was recorded. The same procedure was repeated for several identical brand new relays, with all actuation times and contact resistance data recorded and sorted by actuation count. All data was normalized.

[0067] The above normalized data is used as a training set to input the life prediction model for training. After the training, the relationship between the number of operations, operation time and contact resistance is obtained by fitting the curve according to the output of the life prediction neural network. After testing the relay, the operation time and contact resistance are substituted into the relationship to calculate the theoretical number of operations, thereby predicting the remaining life of the relay.

[0068] The steps in the present invention can be adjusted in sequence, combined, or deleted according to actual needs.

[0069] The units in the device of the present invention can be combined, divided and deleted according to actual needs.

[0070] Although the present invention has been disclosed in detail with reference to the accompanying drawings, it should be understood that these descriptions are merely illustrative and are not intended to limit the application of the present invention. The scope of the present invention is defined by the appended claims and includes various modifications, variations, and equivalents made to the invention without departing from the scope and spirit of the present invention.

Claims

1. A portable relay calibration device, characterized in that: The device includes: a battery pack, an inverter, an industrial computer, a multimeter, a switching power supply, a switch, a switching device, a relay interface and a battery charger; The battery pack converts DC power into AC power through an inverter and inputs it into the power supply port of the industrial computer, the power supply port of the multimeter, the input port of the switching power supply, and the power supply port of the switch. The battery pack powers the industrial computer, the multimeter, and the switch, and provides trigger current for the switching power supply. After the input port of the switching power supply receives the alternating current as the trigger current, the switching power supply provides direct current to the switching device through the output port; the measurement port of the multimeter is connected to the switching device; the switching device is connected to the terminals of the relay interface, and the relay interface is connected to the relay; the switching device is used to bridge the measurement port of the multimeter and the terminals of the relay interface; The industrial computer is connected to the multimeter, switching power supply and switching device through the switch to achieve network communication; The multimeter has three measurement ports: a first measurement port, a second measurement port, and a third measurement port, which are connected to different input ports of a switching device; the relay interface includes three terminals: a coil terminal, a normally closed contact terminal, and a normally open contact terminal, which are respectively connected to different output ports of the switching device; within the switching device, the multimeter connects the first measurement port, the second measurement port, and the third measurement port to the coil terminal, the normally closed contact terminal, and the normally open contact terminal of the relay interface according to preset rules based on specific instructions received; The method of using the device includes: The relay is connected to the relay interface of the switching device. The industrial computer sends a command to the multimeter. The multimeter switches the channels of the first measurement port, the second measurement port, and the third measurement port to the resistance measurement position. The industrial computer also sends a command to the switching device. The switching device connects the first measurement port to the coil terminal and the second measurement port to the normally closed contact terminal. The channel of the multimeter where the first measurement port is located measures the coil resistance of the relay, and the channel where the second measurement port is located measures the normally closed contact resistance of the relay. The multimeter sends the measurement data to the industrial computer. After receiving the measurement data, the industrial computer sends a command to the multimeter to switch the channel where the first measurement port is located to the voltage measurement position. The switching device introduces the DC power output by the switching power supply into the first measurement port. When the channel where the second measurement port of the multimeter is located detects the actuation of the relay contact, that is, the resistance value of the normally closed contact becomes infinite, the voltage measured at the channel where the first measurement port of the multimeter is located is sent to the industrial computer as the actuation voltage of the relay. The switching power supply continues to gradually increase its output voltage. When the output voltage of the switching power supply reaches the rated voltage of the relay and stops increasing, the industrial computer sends a command to the switching device to connect the third measuring port to the normally open contact terminal. The channel corresponding to the third measuring port of the multimeter measures the normally open contact resistance and sends the result to the industrial computer. The multimeter calculates the relay actuation time based on the time from the third measuring port being connected to the normally open contact terminal until the measured resistance stabilizes. The switching power supply gradually reduces its output voltage. During this process, the channel where the third measurement port of the multimeter is located detects the action of the relay contact. That is, when the resistance of the normally open contact becomes infinite, the voltage measured by the channel where the first measurement port of the multimeter is located at this time is sent to the industrial computer as the release voltage of the relay. After completing the above measurements, the host computer displays the test verification results through a standard report and enters the test results into the relay database for subsequent statistical analysis.

2. The portable relay calibration device according to claim 1, wherein: The relay verification device includes five relay interfaces, the same terminal of each relay interface is connected to the same port of the switching device, and each interface performs testing independently.

3. The portable relay testing device according to claim 1, wherein: When the relay verification device is charging, the battery pack is connected to the charging port through the battery charger to supplement the electric energy.

4. A relay life prediction method using the portable relay calibration device according to any one of claims 1 to 3, characterized in that: The method comprises: A lifespan prediction model was established, which consists of an input layer, a TCN layer, an LSTM layer, an attention layer, and an output layer. The TCN layer uses TCN as a feature extractor. TCN treats the sequence data input from the input layer as a one-dimensional signal, performs convolution operations through multiple layers of convolution kernels, and finally inputs the output results into the LSTM layer. The LSTM layer consists of three layers of long short-term memory networks. Each layer contains several time steps. Each time step contains an input gate, a forget gate, and an output gate. The input gate determines how the data of the current time step affects the state of the memory unit; the forget gate determines how the current time step forgets the information of the previous time step; and the output gate determines how the state of the memory unit of the current time step affects the output feature. The first LSTM layer receives the output from the TCN layer, captures the long-term dependencies in the input data, and extracts preliminary temporal features. The processed features are passed to the second and third LSTM layers for further extraction and finally to the attention layer. The output of the LSTM layer is input into the attention layer, which includes the attention mechanism ECA and a fully connected layer. The structure of the ECA module is relatively simple, consisting of three main steps: global average pooling, one-dimensional convolution, and channel weighting. Global average pooling calculates the global average of each channel to obtain a vector representing the importance of the channel. The one-dimensional convolution operation processes this vector and learns the relationship between channels. The fully connected layer connects the output of the attention mechanism and weights the feature map of each channel using the corresponding attention weight. The final result is used as the output of the life prediction model. By repeatedly using the relay, the relationship between the number of relay operations and life is tested; The method for using the relay includes: when it is necessary to calibrate the relay connected to the relay interface, the industrial computer sends a command to the multimeter, the multimeter switches the channels where the first measurement port, the second measurement port, and the third measurement port are located to the resistance measurement gear, the industrial computer simultaneously sends a command to the switching device, the switching device connects the first measurement port to the coil terminal and the second measurement port to the normally closed contact terminal, the channel where the first measurement port of the multimeter is located measures the coil resistance of the relay, and the channel where the second measurement port is located measures the normally closed contact resistance of the relay, and the multimeter sends the measurement data to the industrial computer; After receiving the measurement data, the industrial computer sends a command to the multimeter to switch the channel where the first measurement port is located to the voltage measurement position. The switching device introduces the DC power output by the switching power supply into the first measurement port. When the channel where the second measurement port of the multimeter is located detects the actuation of the relay contact, the resistance of the normally closed contact becomes infinite. The voltage measured by the channel where the first measurement port of the multimeter is sent to the industrial computer as the actuation voltage of the relay. The switching power supply continues to gradually increase its output voltage. When the output voltage of the switching power supply reaches the rated voltage of the relay and stops increasing, the industrial computer sends a command to the switching device to connect the third measuring port to the normally open contact terminal. The channel corresponding to the third measuring port of the multimeter measures the normally open contact resistance and sends the result to the industrial computer. The multimeter calculates the relay actuation time based on the time from the third measuring port being connected to the normally open contact terminal until the measured resistance stabilizes. The switching power supply gradually reduces its output voltage. During this process, the channel where the third measurement port of the multimeter is located detects the action of the relay contact, which is manifested as the resistance of the normally open contact becoming infinite. The voltage measured by the channel where the first measurement port of the multimeter is located at this time is sent to the industrial computer as the release voltage of the relay; After completing the above measurements, the host computer displays the test results through a standard report and enters the test results into the relay database for subsequent statistical analysis; The relay's operating time and contact resistance will increase as the number of relay operations increases. Connect a brand new relay to the portable relay calibration device. Record the operating time and contact resistance every 100 times the relay operates until the relay breaks. After the relay breaks, record the number of operations. Repeat the above operation for several identical brand new relays, record all the operating time and contact resistance data, and sort them in order of the number of relay operations. Normalize all the data. The above normalized data is used as a training set to input the life prediction model for training. After the training, the relationship between the number of operations, operation time and contact resistance is obtained by fitting the curve according to the output of the life prediction neural network. After testing the relay, the operation time and contact resistance are substituted into the relationship to calculate the theoretical number of operations, thereby predicting the remaining life of the relay.

5. The relay life prediction method according to claim 4, wherein: The expression of the convolution operation of the TCN layer is: ,in For input, Indicates the The parallel layer The output of the convolution kernel, Indicates the size , the expansion rate is The convolution kernel, Represents the causal dilated convolution operation.

6. The relay life prediction method according to claim 4, wherein: The life prediction model uses Adam as the optimizer, the learning rate is set to 0.0015, and As the loss function, represents the number of training samples, represents the true value of the sample, Represents the output value of the network.

Images