Equipment automatic testing and synchronous control method and system based on multi-sensor fusion

Abstract

The invention relates to the technical field of power equipment automation test and control, and provides an equipment automation test and synchronous control method and system based on multi-sensor fusion, and the method comprises the steps: obtaining the multi-sensor and operation state data of a switch cabinet, constructing a first training sample set, and training a signal synchronization model; performing feature grouping on each operation state of the equipment by using the model to obtain feature types; constructing a second training sample set based on the feature types and the control data between the two states, and training a synchronous control model; and inputting multi-sensor data of the to-be-tested equipment into the signal synchronization model to obtain characteristic representation, and outputting a synchronization control result through the synchronization control model. According to the method, accurate feature extraction and synchronous control decision of the equipment state are realized through two-stage modeling, a second sample set emphatically selects check point data of which the operation state is stably converted, and the control effect is analyzed and optimized by combining control influence factors.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic testing and control of power equipment. More specifically, the present invention relates to a method and system for automatic testing and synchronous control of equipment based on multi-sensor fusion. Background Art

[0002] In the field of industrial automation, the automatic testing and synchronous control of equipment have always been one of the key technologies. Existing automatic testing methods for equipment mostly rely on single-sensor data acquisition. In the face of complex working conditions, this method often fails to comprehensively and accurately reflect the actual operating state of the equipment. For example, in multi-motor synchronous control, due to the simple control structure of traditional master control and master-slave control methods, the operating states of each branch cannot interact, and the synchronous performance completely depends on the following ability and anti-disturbance ability of each branch. Therefore, the overall synchronous ability of the system is weak. In addition, in the aspect of equipment synchronous control in the prior art, control models based on fixed parameters are mostly used, which are difficult to adapt to the dynamic changes of the equipment operating state.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: First, the single-sensor data acquisition method cannot effectively fuse multi-source information, resulting in incomplete perception of the equipment state and affecting the accuracy and reliability of automatic testing. Second, the existing synchronous control model lacks the adaptive ability to the dynamic changes of the equipment and is difficult to achieve precise synchronous control. Finally, in the aspect of checking the transition of the equipment operating state to stability in the prior art, there is a lack of effective feature extraction and analysis methods, and the stable state of the equipment cannot be accurately identified, thereby affecting the implementation of synchronous control. Summary of the Invention

[0004] The present invention provides a method and system for automatic testing and synchronous control of equipment based on multi-sensor fusion.

[0005] In the first aspect of the present invention, a method for automatic testing and synchronous control of equipment based on multi-sensor fusion is provided, including: Obtaining multi-sensor data of an automatic point-switching switchgear and equipment operating state data, and constructing a first training sample set; constructing a signal synchronization model, and training the signal synchronization model based on the first training sample set to obtain a trained signal synchronization model; Based on the trained signal synchronization model, performing feature grouping on each operating state of the equipment to obtain the feature types of each operating state; Constructing a second training sample set based on the feature types of each operating state and the control data of the equipment between two operating states; constructing a synchronous control model, and training the synchronous control model based on the second training sample set to obtain a trained synchronous control model; Input the multi-sensor data of the device to be tested into the signal synchronization model to obtain the feature representation of the device to be tested; based on the feature representation, obtain the synchronization control result of the device to be tested based on the trained synchronization control model.

[0006] Further, construct a second training sample set based on the feature type of each running state and the control data of the device between two running states, including: Select the check for the transition of the running state to stability, use the feature representation of this stability transition check as the input data of the sample, and use the control data of the device between this stability transition check and the next check as the label of the sample to construct the second training sample set.

[0007] Further, select the check for the transition of the running state to stability in the following way: For each device, calculate the control influence factor between two adjacent running states; If the control influence factor between the t-th running state and the (t + 1)-th running state is positive, then the t-th running state is the check for the transition of the running state to stability.

[0008] Further, calculate the control influence factor between the t-th running state and the (t + 1)-th running state in the following way: Take the t-th running state as the current state and traverse forward in turn until finding a current state with a different feature type from a previous state as the starting state; Take the (t + 1)-th running state as the current state and traverse backward in turn until finding a current state with a different feature type from a subsequent state as the ending state; Calculate the control influence factor between the t-th running state and the (t + 1)-th running state based on the feature types and state times of the starting state and the ending state.

[0009] Further, calculate the control influence factor between the t-th running state and the (t + 1)-th running state based on the feature types and state times of the starting state and the ending state using the following formula: Among them, the feature type of the starting state is the i-th feature type, and the feature type of the ending state is the j-th feature type, represents the distance between the feature type of the starting state and the feature type of the ending state, represents the time of the starting state, represents the time of the ending state, represents the hyperbolic tangent function, represents the sign function, and represents the parameter, Represents the control influence factor between the t-th running state and the (t + 1)-th running state.

[0010] Furthermore, based on the feature representation, obtain the synchronization control result of the device to be tested from the trained synchronization control model, including: Search for K samples similar to the feature representation of the device to be tested in the second training sample set; Obtain the synchronization control results of the K samples based on the trained synchronization control model; Comprehensively evaluate the synchronization control results of the K samples based on the control influence factors of the K samples to obtain the synchronization control result of the device to be tested.

[0011] Furthermore, based on the trained signal synchronization model, perform feature grouping on each running state of the device to obtain the feature type of each running state, including: For the t-th running state of the device, input the running state data of the t-th running state and the previous running states into the trained signal synchronization model to obtain the feature representation of the t-th running state of the device and the signal synchronization prediction result; Perform feature grouping on the feature representations of all running states of all devices to obtain multiple feature types; For each feature type, calculate the average value of the signal synchronization prediction results corresponding to the feature representations belonging to this feature type, and sort the feature types according to the average value of the signal synchronization prediction results to obtain the serial number of each feature type; The feature type of each running state is the feature type to which the feature representation of this running state belongs.

[0012] Furthermore, the signal synchronization model includes: Device inherent parameter extraction module, used to extract inherent attribute features from the device inherent parameters of the samples; Dynamic change feature extraction module, used to extract dynamic change features from the running state data of the samples; Feature fusion module, used to fuse the inherent attribute features and the dynamic change features; Prediction module, used to perform signal synchronization prediction based on the fused features.

[0013] In the second aspect of the present invention, a device automation test and synchronization control system based on multi-sensor fusion is provided, including: Signal synchronization model training module, used to obtain the multi-sensor data of the automatic alignment switch cabinet and the device running state data, construct the first training sample set; construct the signal synchronization model, and train the signal synchronization model based on the first training sample set to obtain the trained signal synchronization model; A feature grouping module, configured to perform feature grouping on the operating state of the device each time based on the trained signal synchronization model, and obtain the feature types of the operating state each time; A synchronization control model training module, configured to construct a second training sample set based on the feature types of the operating state each time and the control data of the device between two operating states; construct a synchronization control model, and train the synchronization control model based on the second training sample set to obtain a trained synchronization control model; A synchronization control module, configured to input the multi-sensor data of the device to be tested into the signal synchronization model to obtain the feature representation of the device to be tested; and obtain the synchronization control result of the device to be tested based on the feature representation and the trained synchronization control model.

[0014] Further, constructing a second training sample set based on the feature types of the operating state each time and the control data of the device between two operating states includes: Selecting the inspection of the transition from the operating state to stability, using the feature representation of this stability transition inspection as the input data of the sample, and using the control data of the device between this stability transition inspection and the next inspection as the label of the sample to construct a second training sample set.

[0015] According to the above embodiments of the present invention, there are at least the following beneficial effects: The device automation test and synchronization control method and system based on multi-sensor fusion can effectively improve the test efficiency and control accuracy of the device. By constructing a signal synchronization model and a synchronization control model, and using multi-sensor data and device operating state data for training, it is possible to accurately perform feature grouping and synchronization control on the device operating state, so that the device can respond more accurately to control instructions during the automated test process, reducing test errors and device failure risks caused by unstable device states or inaccurate control, and ensuring the stability and reliability of device operation. At the same time, this method and system can realize efficient automated test and synchronization control of the device, reducing the cost of manual intervention and test time. During the operation of the device, it can quickly and accurately obtain the feature representation of the device and perform synchronization control, improving the production efficiency and operation efficiency of the device, and saving human and material resources for the enterprise. Description of the Drawings

[0016] By referring to the accompanying drawings and reading the following detailed description, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become easily understandable. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, wherein: Figure 1 It is a schematic flowchart of a device automation test and synchronization control method based on multi-sensor fusion provided by an embodiment of the present invention; Figure 2Schematic diagram of the structure of the device automation test and synchronization control system based on multi-sensor fusion provided by an embodiment of the present invention; Figure 3 Schematically shows a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed implementation manners

[0017] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and then implement the present invention, rather than limiting the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to be able to convey the scope of the present invention completely to those skilled in the art.

[0018] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, a device, an equipment, a method or a computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0019] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0020] Embodiment 1 The following reference is made to Figure 1 , Figure 1 which is a flowchart of the device automation test and synchronization control method based on multi-sensor fusion provided by an embodiment of the present invention. As Figure 1 shown, a device automation test and synchronization control method based on multi-sensor fusion includes: S1 Obtain multi-sensor data and device operation status data of the automatic point-switching switchgear, and construct a first training sample set; construct a signal synchronization model, and train the signal synchronization model based on the first training sample set to obtain a trained signal synchronization model; S2 Based on the trained signal synchronization model, perform feature grouping on each device operation status to obtain the feature types of each operation status; S3 Based on the feature types of each operation status and the control data of the device between two operation statuses, construct a second training sample set; construct a synchronization control model, and train the synchronization control model based on the second training sample set to obtain a trained synchronization control model; S4 Input the multi-sensor data of the device to be tested into the signal synchronization model to obtain the feature representation of the device to be tested; based on the feature representation, obtain the synchronization control result of the device to be tested based on the trained synchronization control model.

[0021] It should be noted that the present invention proposes a method for automatic testing and synchronous control of equipment based on multi-sensor fusion, aiming to achieve automatic testing and precise synchronous control of equipment through the fusion of multi-sensor data and equipment operation status data. Here, the multi-sensor data refers to the data collected from different types of sensors (such as temperature sensors, pressure sensors, current sensors, etc.), which can comprehensively reflect various status information of the equipment during operation. And the equipment operation status data refers to various parameter data generated during the operation of the equipment, such as the operation speed of the equipment, the load condition, etc. By constructing a training sample set and training a signal synchronization model and a synchronous control model, the feature grouping and synchronous control of the equipment operation status can be realized, thereby improving the efficiency and accuracy of automatic equipment testing.

[0022] Specifically, the first training sample set refers to a data set formed by integrating the multi-sensor data and equipment operation status data of an automatic point-switching switchgear, which is used to train the signal synchronization model. The signal synchronization model is a model for predicting the signal synchronization situation of the equipment. It analyzes the input multi-sensor data and equipment operation status data and outputs the signal synchronization prediction result of the equipment. Feature grouping means classifying the features of each equipment operation status according to the output of the signal synchronization model to obtain different types of feature representations. The second training sample set is constructed based on the feature types of each operation status and the control data of the equipment between two operation statuses, and is used to train the synchronous control model. The synchronous control model can output the synchronous control result of the equipment according to the input feature representation.

[0023] The first training sample set is formed by collecting the multi-sensor data (such as data collected by temperature, pressure, current, etc. sensors) and equipment operation status data (such as parameter data such as the operation speed and load condition of the equipment) of the automatic point-switching switchgear in different operation states, and integrating these data together. For example, the multi-sensor data and operation status data of a certain model of automatic point-switching switchgear in different states such as normal operation, overload operation, and fault operation are collected. Each group of data is used as a sample, including multiple feature dimensions such as temperature value, pressure value, current value, operation speed value, etc., so as to construct the first training sample set for training the signal synchronization model.

[0024] Preferably, when constructing the signal synchronization model, machine learning algorithms such as neural networks or support vector machines can be used. The input parameters include multi-sensor data and device operating status data, which, after preprocessing, serve as the input features of the model. When calculating the control influence factor, the change trend of the device status can be evaluated by analyzing the feature types and time information between two adjacent operating states. For example, if the control influence factor between the t-th operating state and the (t + 1)-th operating state is positive, it indicates that the device is changing towards a stable direction between these two operating states. During the training process of the synchronization control model, a supervised learning method can be adopted to optimize the model parameters by minimizing the error between the prediction result and the actual control data. When obtaining the synchronization control result of the device to be tested, based on the feature representation of the device to be tested, similar samples can be searched in the second training sample set, and the control influence factors of these samples can be combined for comprehensive evaluation to obtain a more accurate synchronization control result.

[0025] The second training sample set is constructed based on the feature types of each operating state and the control data of the device between two operating states. First, the checkpoints where the operating state changes towards stability are determined. The feature representations of these checkpoints are used as the input data of the samples, and the control data of the device from this checkpoint to the next checkpoint is used as the label of the sample. For example, during the process of the device changing from the overload state to the normal operating state, several key checkpoints are selected, and the feature representations of these checkpoints (such as the combination of feature parameters like the temperature sensor reading starting to decline from a high value and the current sensor reading gradually stabilizing) and the control data executed during the period from this checkpoint to the device running stably completely (such as the control instruction to reduce the load, the control parameter to adjust the motor speed, etc.) are recorded. They are combined into a sample, and multiple such samples form the second training sample set for training the synchronization control model.

[0026] The signal synchronization model includes a device inherent parameter extraction module, a dynamic change feature extraction module, a feature fusion module, and a prediction module.

[0027] Device inherent parameter extraction module: Extract inherent attribute features from the device inherent parameters of the sample (such as the model, rated power, rated current, etc. of the device). For example, for automatic point-switching switch cabinets of different models, their rated power, rated current, etc. are different, and these parameters remain unchanged during the device operation, reflecting the basic characteristics of the device. Through this module, these inherent parameters can be extracted as part of the features.

[0028] Dynamic change feature extraction module: Extract dynamic change features from the running state data of samples (such as real-time data like temperature, pressure, current, voltage during device operation). For example, during device operation, a temperature sensor monitors the temperature change of the device in real time, and a current sensor monitors the current fluctuation. These dynamic change data reflect the real-time state of the device during operation. Through this module, dynamic features such as temperature change rate and current fluctuation amplitude can be extracted.

[0029] Feature fusion module: Fuse the inherent attribute features and dynamic change features. For example, weightedly fuse the rated power of the device with dynamic features such as current and voltage during current operation to form a comprehensive feature vector, achieving a comprehensive characterization of the device running state.

[0030] Prediction module: Perform signal synchronization prediction based on the fused features. For example, a neural network algorithm can be used. Taking the fused feature vector as input, train the neural network model so that it can predict the signal synchronization situation of the device in the current running state, and output the signal synchronization prediction result of the device, such as whether the device is in a synchronous state and the synchronization accuracy.

[0031] Train the signal synchronization model using the first training sample set. Input the multi-sensor data and device running state data in the sample into the model, and adjust the parameters of the model to make the prediction result of the model as close as possible to the actual situation in the sample. Through continuous iterative training, finally obtain the trained signal synchronization model.

[0032] In some embodiments, construct a second training sample set based on the feature type of each running state and the control data of the device between two running states, including: Select the inspection of the transition of the running state to stability. Use the feature representation of this stability transition inspection as the input data of the sample, and the control data of the device between this stability transition inspection and the next inspection as the label of the sample to construct the second training sample set.

[0033] It should be noted that when constructing the second training sample set in the present invention, a specific sample selection strategy is adopted, that is, the inspection of the transition of the operating state to stability is selected as the input data of the sample, and the control data of the device between this stability transition inspection and the next inspection is used as the label of the sample. The core of this strategy lies in capturing the relationship between the feature representation and the control data of the device at this critical stage by analyzing the stable transition process of the device operating state, so as to provide more representative and targeted training samples for the synchronous control model. Among them, the inspection of the transition of the operating state to stability refers to the process in which the device gradually transitions from an unstable state to a stable operating state. The feature representation in this process can reflect the key features of the device during the stable process, and the control data refers to the relevant data for controlling the device during this process, such as the parameters of the control instruction, the control frequency, etc. These data are used as the labels of the samples to guide the learning of the synchronous control model.

[0034] Specifically, the second training sample set refers to the data set used to train the synchronous control model, which consists of input data and corresponding labels. The input data is the feature representation of the transition of the device operating state to stability, and these feature representations can include the change trend of sensor data and the adjustment of operating parameters during the stable transition process of the device. The label is the control data of the device between this stability transition inspection and the next inspection, and these control data can be specific control instruction values, the execution time of control actions, etc. For example, if during the stable transition process of the device, the reading of the temperature sensor gradually decreases from a high temperature to the normal operating temperature range, and at the same time the rotational speed of the device gradually adjusts from a high speed to a stable operating speed, these change trends and adjustment parameters are part of the feature representation, and the cooling instruction, speed adjustment instruction and their execution time issued during this process are the corresponding control data.

[0035] In some embodiments, the second training sample set is constructed based on the feature type of each operating state and the control data of the device between two operating states. First, determine the inspection points of the transition of the operating state to stability, use the feature representations of these inspection points as the input data of the sample, and use the control data of the device between this inspection point and the next inspection as the label of the sample. For example, during the transition of the device from an overload state to a normal operating state, select several key inspection points, record the feature representations of these inspection points (such as the combination of feature parameters such as the reading of the temperature sensor starting to decrease from an excessive value and the reading of the current sensor gradually stabilizing), and the control data executed during the period from this inspection point to the device operating stably (such as the control instruction to reduce the load, the control parameters to adjust the motor speed, etc.), and form a sample with them. Multiple such samples constitute the second training sample set for training the synchronous control model.

[0036] Preferably, when selecting the inspection of the transition from the selection running state to stability, a certain threshold or condition can be set to determine whether the device is in a stable transition state. For example, conditions such as the reading of the temperature sensor dropping to a certain range and the vibration amplitude of the device being lower than a certain threshold can be set. When these conditions are met, the device is considered to be in a stable transition state, and the feature representation and subsequent control data at this time are used as samples. During the process of constructing the samples, preprocessing operations such as normalization and dimensionality reduction can also be performed on the feature representation to improve the training effect of the model. For the control data, it can be classified and encoded according to the type and execution frequency of the control instructions to make it more suitable as the label for model training. Through these refined operation steps, the second training sample set can be constructed more accurately, thereby improving the performance and reliability of the synchronous control model.

[0037] In some embodiments, the following method is adopted to select the inspection of the transition from the running state to stability: For each device, calculate the control influence factor between two adjacent running states; If the control influence factor between the t-th running state and the (t + 1)-th running state is positive, then the t-th running state is the inspection of the transition from the running state to stability.

[0038] It should be noted that the present invention selects the inspection of the transition from the running state to stability by calculating the control influence factor between two adjacent running states. The control influence factor is an index that measures the change trend of the device running state, and it reflects the influence of the control operation between two adjacent running states of the device on the stability of the device. If the control influence factor is positive, it means that the device is transitioning from an unstable state to a stable state, and this transition process is of great significance for constructing the second training sample set because it can help screen out the key features and control data of the device during the stable transition process.

[0039] Specifically, the control influence factor is a parameter that quantifies the change trend of the device running state, and it is calculated by analyzing the feature changes and control operations of the device between two adjacent running states. For example, during the operation of the device, if a control operation causes the temperature of the device to gradually decrease from an excessive state to the normal working range, and the vibration amplitude of the device also decreases accordingly, then the control influence factor of this control operation may be positive, indicating that the device is transitioning to a stable state. When calculating the control influence factor, various running state parameters of the device, such as temperature, pressure, current, etc., and the corresponding control operation parameters, such as the intensity and execution time of the control instruction, need to be considered. The specific values of these parameters can be obtained through sensor acquisition and system control records, thereby providing data support for the calculation of the control influence factor.

[0040] Preferably, when calculating the control influence factor, a method based on time series analysis can be adopted. First, determine the starting state and the ending state of the device operation state, and these two states correspond to the start and end time points of the change in the device operation state respectively. Then, according to the characteristic types and state times of the starting state and the ending state, calculate the characteristic change amount of the device during this time period and the influence degree of the control operation. For example, by calculating the change amounts of characteristic parameters such as the temperature change rate and the pressure fluctuation amplitude during the stable transition process of the device, combined with the execution time and intensity of the control instruction, the actual influence of the control operation on the device stability can be evaluated. This method can more accurately reflect the dynamic changes of the device during operation, thus providing a reliable basis for selecting the inspection of the transition from the operation state to stability.

[0041] In some embodiments, the following method is adopted to calculate the control influence factor between the t-th operation state and the (t + 1)-th operation state: Taking the t-th operation state as the current state, traverse forward in turn until a current state with a different characteristic type from a previous state is found as the starting state; Taking the (t + 1)-th operation state as the current state, traverse backward in turn until a current state with a different characteristic type from a subsequent state is found as the ending state; Calculate the control influence factor between the t-th operation state and the (t + 1)-th operation state based on the characteristic types and state times of the starting state and the ending state.

[0042] It should be noted that when calculating the control influence factor between the t-th operation state and the (t + 1)-th operation state in the present invention, a method that comprehensively considers the characteristic types and state times of the starting state and the ending state is adopted. This method traverses the device operation states forward and backward to find the starting state and the ending state where the characteristic types change, and calculates the control influence factor based on the characteristic types, state times and related parameters of these states. The control influence factor is an important index used to evaluate the change trend of the device between two adjacent operation states, especially to judge whether the device changes from an unstable state to a stable state, thus providing a basis for subsequent synchronous control.

[0043] Specifically, the starting state refers to the first state found by traversing forward from the t-th running state, which is different from the current state feature type, and it marks the starting point of the change in the device running state. For example, if the device gradually recovers from an overload state to a normal running state, then the starting state is the overload state. The ending state is the first state found by traversing backward from the (t + 1)-th running state, which is different from the feature type of the subsequent state, representing the end point of the change in the device running state. For example, if the device enters a fault state from a normal running state, then the ending state is the fault state. The feature type is the result of classifying according to the feature representation of the device running state. For example, the device running state can be classified into different types such as normal running, overload running, fault running, etc., and each type corresponds to a different feature representation. The state time refers to the time point or time period when the device is in a certain running state, used to measure the time span of the state change. When calculating the control influence factor, it is necessary to consider the distance between the feature types of the starting state and the ending state, which reflects the difference degree between the two states in the feature space. In addition, the time factor also needs to be considered, that is, the difference between the ending state time and the starting state time, used to evaluate the influence of the time span of the state change on the control influence factor.

[0044] Preferably, when calculating the control influence factor, a weighted method can be adopted to comprehensively consider the influence of the distance between feature types, state time, and the sign function. First, according to the distance between feature types and state time, a basic influence factor is calculated, and this factor is normalized by the hyperbolic tangent function to ensure its value is within a certain range. The role of the hyperbolic tangent function is to map the ratio of the distance between feature types and state time to a fixed interval, thereby avoiding the influence of too large or too small values on the calculation result. Then, according to the relative relationship between feature types, the sign of the influence factor is adjusted by the sign function to reflect whether the device running state changes towards a stable direction or an unstable direction. The role of the sign function is to judge the change trend of the device state according to the positive or negative of the distance between feature types. Finally, the influence of the basic influence factor and the sign function is weighted by two parameters to obtain the final control influence factor. The specific values of these two parameters can be adjusted according to the actual operating characteristics and control requirements of the device to ensure that the control influence factor can accurately reflect the change trend of the device running state.

[0045] For each operating state of the device, input the operating state data of the t-th time and the previous operating states into the trained signal synchronization model to obtain the feature representation of the device's t-th operating state and the signal synchronization prediction result. Group the feature representations of all operating states of all devices to obtain multiple feature types. For example, by processing the data of the device in different operating states through the signal synchronization model, the feature vector of each operating state is obtained, and then the clustering algorithm (such as K-means clustering) is used to perform clustering analysis on these feature vectors. The similar feature vectors are divided into the same group, and each group corresponds to a feature type, such as the normal operating feature type, the light overload feature type, the severe overload feature type, etc.

[0046] For each feature type, calculate the average value of the signal synchronization prediction results corresponding to the feature representations belonging to this feature type, and sort the feature types according to the average value of the signal synchronization prediction results to obtain the serial number of each feature type. For example, the average value of the signal synchronization prediction results of the normal operating feature type may be higher, while the average value of the signal synchronization prediction results of the severe overload feature type may be lower. Sorting the feature types according to this average value can determine the importance and priority of each feature type in the device operating state, so as to be able to process them targeted in subsequent synchronization control.

[0047] In some embodiments, based on the feature types and state times of the starting state and the ending state, the following formula is used to calculate the control influence factor between the t-th operating state and the (t + 1)-th operating state: where the feature type of the starting state is the i-th feature type, and the feature type of the ending state is the j-th feature type, represents the distance between the feature type of the starting state and the feature type of the ending state, represents the time of the starting state, represents the time of the ending state, represents the hyperbolic tangent function, represents the sign function, and represent parameters, represents the control influence factor between the t-th operating state and the (t + 1)-th operating state.

[0048] It should be noted that the calculation formula of the control influence factor mentioned in the present invention is a model that comprehensively considers various factors and is used to quantify the change trend between the operating states of the device. Through the weighted combination of the distance between feature types, the state time, and relevant parameters, this formula can more accurately evaluate the control effect when the device changes from one operating state to another. This method is particularly suitable for the automated testing and synchronous control scenarios of complex devices and can help the system more accurately identify the stable transition process of the device.

[0049] The device operating state data is shown in Table 1: Table 1 Calculation process: Suppose we want to calculate the control influence factor between the 3rd operating state and the 4th operating state, and = 0.6, = 0.4.

[0050] Determine the starting state and the ending state: Taking the 3rd operating state as the current state, traverse forward. The feature types of the 3rd operating state are 50°C and 1.4 MPa, the 2nd is 48°C and 1.3 MPa, and the 1st is 45°C and 1.2 MPa. Assuming that the feature types are distinguished by the combination of temperature and pressure, it is found that the feature types of the 3rd time are different from the previous two times (both temperature and pressure are rising), so the starting state is the 3rd operating state.

[0051] Taking the 4th operating state as the current state, traverse backward. The feature types of the 4th operating state are 47°C and 1.25 MPa, and the 5th is 46°C and 1.15 MPa. It is found that the feature types of the 4th time are different from the 5th time (both temperature and pressure are falling), so the ending state is the 4th operating state.

[0052] Calculate the feature type distance: Suppose the feature type distance is calculated using the Euclidean distance. The feature vector of the starting state (the 3rd time) is (50, 1.4), and the feature vector of the ending state (the 4th time) is (47, 1.25).

[0053] The distance is: Calculate the time difference: The starting state time is 30 seconds, the ending state time is 40 seconds, and the time difference is 40 - 30 = 10 seconds Substitute into the formula to calculate the control influence factor: Assume that the sign function is determined to be positive based on the changing trend of the device state (because both the temperature and pressure are decreasing from the 3rd to the 4th time, indicating a transition towards a stable state), then: It can be obtained that: 。

[0054] Specifically, the control influence factor is a quantitative index that measures the changing trend of the device operation state. It is calculated by comprehensively considering the distance between the characteristic types of the starting state and the ending state, the state time, and relevant parameters. Among them, the characteristic type of the starting state refers to the characteristic type of the device when the change starts. For example, when the device gradually recovers from an overload state to a normal operation state, the characteristic type of the starting state is the overload state. The characteristic type of the ending state refers to the characteristic type of the device when the change ends. For example, when the device enters a fault state from a normal operation state, the characteristic type of the ending state is the fault state. The distance between the characteristic types refers to the degree of difference between the characteristic types of the starting state and the ending state in the characteristic space. For example, if the device gradually cools from a high-temperature state to the normal temperature range, this distance can represent the amplitude of the temperature change. The state time refers to the time experienced by the device from the starting state to the ending state. For example, the time required for the device to recover from an overload state to a normal operation state. The parameters α and β in the formula are coefficients used to adjust the weights of different factors and can be adjusted according to the specific characteristics and control requirements of the device.

[0055] Preferably, when calculating the control influence factor, the various parameters and steps in the formula can be further refined. For example, the distance between the characteristic types can be determined by calculating the Euclidean distance between the characteristic vectors of the two states, which can reflect the characteristic differences of the device in different states. The state time can be obtained from the timestamps in the device operation log to ensure the accuracy of time calculation. The parameters α and β can be optimized according to the operation characteristics and control objectives of the device. For example, if the device is sensitive to time, the weight of β can be increased; if the device is sensitive to characteristic differences, the weight of α can be increased. In practical applications, the optimal values of these parameters can be determined through experiments and data analysis to improve the calculation accuracy and reliability of the control influence factor.

[0056] In some embodiments, based on the feature representation, obtaining the synchronization control result of the device to be tested from the trained synchronization control model includes: Searching for K samples similar to the feature representation of the device to be tested in the second training sample set; Obtaining the synchronization control results of the K samples based on the trained synchronization control model; Based on the control influence factors of K samples, comprehensively evaluate the synchronous control results of the K samples to obtain the synchronous control results of the device under test.

[0057] It should be noted that when obtaining the synchronous control results of the device under test according to the feature representation in the present invention, a method based on comprehensive evaluation of similar samples is adopted. This method finds samples similar to the feature representation of the device under test in the second training sample set, and comprehensively evaluates by combining the synchronous control results of these samples and their control influence factors, so as to obtain the synchronous control results of the device under test. The core of this method lies in using the information in the existing training sample set, through similarity matching and comprehensive evaluation, to improve the accuracy and reliability of the synchronous control results.

[0058] Specifically, the second training sample set refers to the sample set used when constructing the synchronous control model, which contains the feature representation of the device operating state and its corresponding synchronous control results. These samples are obtained by analyzing the process of the device operating state changing to stability, and have high representativeness and guiding significance. The feature representation of the device under test refers to the feature vector of the device operating state obtained through the signal synchronization model, which can reflect the main features of the device in the current operating state. Similar samples refer to the samples in the second training sample set that are similar to the feature representation of the device under test, and this similarity can be measured by calculating the distance or similarity between feature vectors. The control influence factor is an index to measure the control effect of a sample, which reflects the importance and reliability of the sample in the synchronous control process. By finding similar samples and comprehensively evaluating in combination with their control influence factors, the synchronous control results of the device under test can be obtained.

[0059] Preferably, in the implementation process, the following steps can be adopted to refine the operation. First, define a similarity measurement method, such as using Euclidean distance or cosine similarity to calculate the similarity between the feature representation of the device under test and the feature representation of each sample in the second training sample set. Then, according to the set similarity threshold, screen out K samples similar to the feature representation of the device under test from the second training sample set. Next, for these K similar samples, extract their synchronous control results and control influence factors. Finally, according to the weights of the control influence factors, perform weighted averaging on the synchronous control results of these K samples to obtain the final synchronous control results of the device under test. This method can make full use of the information in the training sample set, improve the accuracy and reliability of the synchronous control results, and at the same time, the similarity threshold and weight parameters can be adjusted according to actual needs to adapt to different devices and control scenarios.

[0060] Input the multi-sensor data of the device under test into the signal synchronization model. The model will process the input data, extract the inherent property features and dynamic change features of the device, and fuse them to finally obtain the feature representation of the device under test. This feature representation is a feature vector that comprehensively reflects the current operating state of the device, containing various information such as the inherent parameters and dynamic operating parameters of the device, and can accurately characterize the operating state of the device at the current moment.

[0061] In some embodiments, search for K samples in the second training sample set that are similar to the feature representation of the device under test. Obtain the synchronization control results of these K samples based on the trained synchronization control model, and then comprehensively evaluate the synchronization control results of the K samples according to the control influence factors of the K samples to obtain the synchronization control result of the device under test. For example, assume that the feature representation of the device under test is similar to the K samples in the second training sample set. These K samples correspond to different synchronization control strategies and control effects respectively. By analyzing the control influence factors of these samples (such as the trend of the control operation to make the device state change towards stability, etc.), use comprehensive evaluation methods such as weighted average for the synchronization control results of the K samples to determine the synchronization control result that is most suitable for the current state of the device under test, such as adjusting the operating parameters of the device, issuing specific control instructions, etc., to achieve precise synchronization control of the device.

[0062] Search for K samples in the second training sample set that are similar to the feature representation of the device under test. Obtain the synchronization control results of these K samples based on the trained synchronization control model, and then comprehensively evaluate the synchronization control results of the K samples according to the control influence factors of the K samples to obtain the synchronization control result of the device under test. For example, assume that the feature representation of the device under test is similar to the K samples in the second training sample set. These K samples correspond to different synchronization control strategies and control effects respectively. By analyzing the control influence factors of these samples (such as the trend of the control operation to make the device state change towards stability, etc.), use comprehensive evaluation methods such as weighted average for the synchronization control results of the K samples to determine the synchronization control result that is most suitable for the current state of the device under test, such as adjusting the operating parameters of the device, issuing specific control instructions, etc., to achieve precise synchronization control of the device.

[0063] In some embodiments, the feature representation of the device under test can also be input into the trained synchronization control model, and the synchronization control model directly outputs the synchronization control result of the device under test. For example, the synchronization control model can predict the control measures to be taken according to the feature representation of the device under test, such as the combination of parameters such as the current temperature, pressure, and current of the device, such as increasing the power of the cooling system, adjusting the speed of the motor, etc., to make the device reach the best state of synchronous operation.

[0064] In some embodiments, the feature representation of the device under test can also be input into the trained synchronization control model, and the synchronization control model directly outputs the synchronization control result of the device under test. For example, based on the feature representation of the device under test, such as the combination of parameters such as the current temperature, pressure, and current of the device, the synchronization control model can predict the control measures to be taken, such as increasing the power of the cooling system, adjusting the rotation speed of the motor, etc., so that the device reaches the optimal state of synchronous operation.

[0065] In some embodiments, based on the trained signal synchronization model, the feature grouping of each operating state of the device is performed to obtain the feature types of each operating state, including: For the t-th operating state of the device, the t-th operating state and the previous operating state data are input into the trained signal synchronization model to obtain the feature representation of the t-th operating state of the device and the signal synchronization prediction result; Feature grouping is performed on the feature representations of all operating states of all devices to obtain multiple feature types; For each feature type, calculate the average value of the signal synchronization prediction results corresponding to the feature representations belonging to the feature type, and sort the feature types according to the average value of the signal synchronization prediction results to obtain the serial number of each feature type; The feature type of each operating state is the feature type to which the feature representation of the operating state belongs.

[0066] It should be noted that when the present invention performs feature grouping on each operating state of the device based on the trained signal synchronization model, a systematic method is adopted. This method analyzes the operating state data of the device, extracts the feature representations of each operating state, and performs grouping based on these feature representations, thereby obtaining different types of feature types. This method can effectively identify the key features of the device in different operating states and provide an accurate basis for subsequent synchronization control. By calculating the average value of the signal synchronization prediction results of each feature type and sorting the feature types according to these average values, the effect of feature grouping can be further optimized, and the accuracy and reliability of the model can be improved.

[0067] Specifically, the signal synchronization model is a trained model used to analyze the multi-sensor data and operating status data of a device and output the signal synchronization prediction result of the device. By fusing the inherent attribute features and dynamic change features of the device, this model can accurately reflect the feature representation of the device under different operating states. Feature representation refers to the feature vector of the device under a certain operating state, which contains the main feature information of the device in this state, such as the change trend of sensor data, the values of operating parameters, etc. Feature grouping means classifying the feature representations of all operating states of the device according to similarity to obtain different types of feature types. Each feature type represents the set of features of the device under a certain specific operating state. The average value of the signal synchronization prediction results refers to the value obtained by averaging the signal synchronization prediction results belonging to the same feature type, which reflects the typical synchronization state of this feature type. The serial number of the feature type is the number obtained by sorting the feature types according to the average value of the signal synchronization prediction results, and is used to identify the importance and priority of different feature types.

[0068] The signal synchronization model includes: The signal synchronization model includes a device inherent parameter extraction module, a dynamic change feature extraction module, a feature fusion module, and a prediction module.

[0069] Device inherent parameter extraction module: Extract inherent attribute features from the device inherent parameters of the sample (such as the model, rated power, rated current, etc. of the device). For example, for automatic point-switching switchgear of different models, their rated power, rated current and other parameters are different, and these parameters remain unchanged during the operation of the device, reflecting the basic characteristics of the device. Through this module, these inherent parameters can be extracted as part of the features.

[0070] Dynamic change feature extraction module: Extract dynamic change features from the operating status data of the sample (such as real-time data such as temperature, pressure, current, voltage, etc. during device operation). For example, during the operation of the device, the temperature sensor will monitor the temperature change of the device in real time, and the current sensor will monitor the current fluctuation. These dynamic change data reflect the real-time state of the device during operation. Through this module, dynamic features such as temperature change rate and current fluctuation amplitude can be extracted.

[0071] Feature fusion module: Fuse the inherent attribute features and dynamic change features. For example, weight and fuse the rated power of the device with dynamic features such as current and voltage during current operation to form a comprehensive feature vector to achieve a comprehensive characterization of the device operating state.

[0072] Prediction module: Based on the fused features, perform signal synchronization prediction. For example, a neural network algorithm can be used, with the fused feature vector as the input, to train a neural network model so that it can predict the signal synchronization situation of the device in the current operating state and output the signal synchronization prediction result of the device, such as whether the device is in a synchronous state, the synchronization accuracy, etc.

[0073] Use the first training sample set to train the signal synchronization model. Input the multi-sensor data and device operating state data in the sample into the model, and adjust the parameters of the model so that the prediction result of the model is as close as possible to the actual situation in the sample. Through continuous iterative training, finally obtain a trained signal synchronization model.

[0074] Preferably, during the implementation process, first, input the operating state of the device each time and its previous operating state data into the trained signal synchronization model to obtain the feature representation and signal synchronization prediction result of each operating state. Then, perform clustering analysis on the feature representations of all operating states of all devices. For example, use the K-means clustering algorithm to divide the operating states with similar feature representations into the same group, thereby obtaining multiple feature types. Next, for each feature type, calculate the average value of all its signal synchronization prediction results, and sort the feature types according to these average values to obtain the serial number of each feature type. Finally, determine the feature type of each operating state according to the feature type to which the feature representation belongs. This method can make full use of the prediction ability of the signal synchronization model, and through clustering analysis and sorting optimization, improve the accuracy and efficiency of feature grouping, and provide more accurate input for the synchronous control of the device.

[0075] For each operating state of the device, input the t-th operating state and its previous operating state data into the trained signal synchronization model to obtain the feature representation and signal synchronization prediction result of the t-th operating state of the device. Perform feature grouping on the feature representations of all operating states of all devices to obtain multiple feature types. For example, process the data of the device in different operating states through the signal synchronization model to obtain the feature vector of each operating state, and then use a clustering algorithm (such as K-means clustering) to perform clustering analysis on these feature vectors, and divide the similar feature vectors into the same group. Each group corresponds to a feature type, such as the normal operating feature type, the light overload feature type, the severe overload feature type, etc.

[0076] For each feature type, calculate the average value of the signal synchronization prediction results corresponding to the feature representations belonging to that feature type, and sort the feature types according to the average value of the signal synchronization prediction results to obtain the serial number of each feature type. For example, the average value of the signal synchronization prediction results of the normal operation feature type may be relatively high, while the average value of the signal synchronization prediction results of the severe overload feature type may be relatively low. Sorting the feature types according to this average value can determine the importance and priority of each feature type in the device operating state, so as to be able to handle it specifically in subsequent synchronization control.

[0077] Input the multi-sensor data of the device to be tested into the signal synchronization model. The model will process the input data, extract the inherent attribute features and dynamic change features of the device, and perform fusion, and finally obtain the feature representation of the device to be tested. This feature representation is a feature vector that comprehensively reflects the current operating state of the device, contains multi-faceted information such as the inherent parameters and dynamic operating parameters of the device, and can accurately characterize the operating state of the device at the current moment.

[0078] In some embodiments, the signal synchronization model includes: An inherent parameter extraction module of the device, configured to extract inherent attribute features from the inherent parameters of the sample; A dynamic change feature extraction module, configured to extract dynamic change features from the operating state data of the sample; A feature fusion module, configured to fuse the inherent attribute features and the dynamic change features; A prediction module, configured to perform signal synchronization prediction based on the fused features.

[0079] It should be noted that the signal synchronization model is one of the core components in the present invention for processing device automated testing and synchronization control. The model extracts the inherent attribute features and dynamic change features of the device, and fuses these features, and finally realizes the prediction of device signal synchronization. This design can effectively integrate the static and dynamic information of the device, provide accurate prediction results for the automated testing and synchronization control of the device, thereby improving the overall performance and reliability of the system.

[0080] Specifically, the signal synchronization model consists of multiple modules, including a device inherent parameter extraction module, a dynamic change feature extraction module, a feature fusion module, and a prediction module. The role of the device inherent parameter extraction module is to extract inherent attribute features from the device inherent parameters of the samples. These inherent parameters may include the device model, rated power, rated current, etc. These parameters remain unchanged during the device operation and reflect the basic characteristics of the device. The dynamic change feature extraction module extracts dynamic change features from the operation status data of the samples. These operation status data may include real-time data such as temperature, pressure, current, and voltage during device operation. These data change with the change of the device operation status and reflect the real-time operation of the device. The feature fusion module fuses the inherent attribute features and dynamic change features to form a comprehensive feature vector for subsequent signal synchronization prediction. The prediction module then performs signal synchronization prediction based on the fused features and outputs the signal synchronization prediction result of the device.

[0081] Preferably, when constructing the signal synchronization model, the following steps can be adopted. First, standardize the device inherent parameters to facilitate subsequent feature fusion. Then, preprocess the dynamic change features, such as removing noise and filling missing values, to improve the quality of the features. Next, use appropriate feature fusion methods, such as weighted summation and feature splicing, to fuse the inherent attribute features and dynamic change features. Finally, select appropriate prediction algorithms, such as neural networks and support vector machines, to construct the prediction module and use the training data to train and optimize the model. In practical applications, the parameters and algorithms of each module can be adjusted according to the specific type and operating environment of the device to achieve the best prediction effect.

[0082] The following is an application embodiment of the present application in combination with a specific scenario. On the production line of a certain factory, the automatic alignment switchgear is one of the key devices, and its main function is to control the distribution and switching of electricity to ensure the stable operation of the production line. However, there are many deficiencies in traditional automated testing and control methods. For example, the single-sensor data acquisition method cannot comprehensively and accurately reflect the actual operation status of the device, and the synchronization control model lacks the adaptive ability to the dynamic changes of the device, making it difficult to achieve precise synchronization control. To solve these problems, a device automation testing and synchronization control method and system based on multi-sensor fusion are adopted.

[0083] First, data is collected in real time from various sensors installed on the automatic point-switching switchgear. These sensors include temperature sensors, pressure sensors, current sensors, voltage sensors, etc. They collect the temperature and pressure of key components inside the equipment, as well as the current and voltage of the equipment at a frequency of once per second respectively. At the same time, the operating parameters of the equipment, such as operating speed, load condition, etc., and the execution status of control instructions are also recorded. For example, the data collected by the temperature sensor may be 45°C, 46°C, 47°C, etc., the data collected by the pressure sensor may be 1.2 MPa, 1.22 MPa, 1.25 MPa, etc., the data collected by the current sensor may be 10 A, 10.2 A, 10.5 A, etc., and the data collected by the voltage sensor may be 380 V, 381 V, 382 V, etc. The operating speed of the equipment may be 1450 r / min, 1460 r / min, 1470 r / min, etc., and the load condition may be 60%, 62%, 65%, etc. These data are preprocessed, including operations such as data cleaning, normalization, and feature engineering, to improve the quality and usability of the data.

[0084] Next, a first training sample set is constructed. 1800 groups of data are selected from the multi-sensor data and equipment operating state data of the equipment in different states such as normal operation, overload operation, and fault operation. Among them, 1000 groups of data are selected in the normal operation state, 500 groups of data are selected in the overload operation state, and 300 groups of data are selected in the fault operation state. Each group of data is used as a sample, containing multiple feature dimensions, such as temperature value, pressure value, current value, voltage value, operating speed value, load percentage, and control instruction execution time, etc. For example, a sample may contain feature values such as temperature value 45°C, pressure value 1.2 MPa, current value 10 A, voltage value 380 V, operating speed value 1450 r / min, load percentage 60%, and control instruction execution time 0.1 s, etc. These samples are used to train the signal synchronization model.

[0085] The signal synchronization model includes an equipment inherent parameter extraction module, a dynamic change feature extraction module, a feature fusion module, and a prediction module. The equipment inherent parameter extraction module extracts inherent attribute features from the equipment inherent parameters of the sample, such as the model, rated power, rated current, etc. of the equipment. The dynamic change feature extraction module extracts dynamic change features from the operating state data of the sample, such as temperature change rate, current fluctuation amplitude, etc. The feature fusion module fuses the inherent attribute features and dynamic change features to form a comprehensive feature vector. The prediction module performs signal synchronization prediction based on the fused features and outputs the signal synchronization prediction result of the equipment. The prediction module is constructed using a neural network algorithm, and the model is trained by setting a loss function and an optimization algorithm. After multiple iterative trainings, a trained signal synchronization model is obtained.

[0086] Then, based on the trained signal synchronization model, the feature grouping of each operating state of the device is performed to obtain the feature types of each operating state. For the t-th operating state of the device, the t-th operating state and the previous operating state data are input into the trained signal synchronization model to obtain the feature representation of the t-th operating state of the device and the signal synchronization prediction result. Cluster analysis is performed on the feature representations of all operating states of all devices, and similar feature representations are divided into the same group, thereby obtaining multiple feature types. For example, a normal operation feature type, a light overload feature type, a severe overload feature type, etc. are obtained. For each feature type, calculate the average value of the signal synchronization prediction results corresponding to the feature representations belonging to this feature type, and sort the feature types according to the average value of the signal synchronization prediction results to obtain the serial number of each feature type. For example, the serial number of the normal operation feature type is 1, the serial number of the light overload feature type is 2, and the serial number of the severe overload feature type is 3.

[0087] Next, a second training sample set is constructed. Checkpoints where the operating state changes to stability are selected, and the feature representations of these checkpoints are used as the input data of the samples, and the control data of the device from this checkpoint to the next check are used as the labels of the samples. For example, during the process of the device changing from the overload state to the normal operation state, 5 key checkpoints are selected. At these checkpoints, the feature representations of the device are recorded, such as the temperature value of 48 °C, the pressure value of 1.3 MPa, the current value of 10.8 A, the voltage value of 385 V, the operating speed value of 1480 r / min, the load percentage of 70%, etc., and the control data executed during the period from this checkpoint to the device running stably completely, such as the control instruction to reduce the load (the control instruction value is to reduce the load by 10%), the control parameter to adjust the motor speed (reduce the motor speed from 1500 r / min to 1450 r / min), etc. The checkpoints where the operating state changes to stability are selected by calculating the control influence factor between two adjacent operating states. For example, for the 3rd operating state and the 4th operating state, calculate the control influence factor between them. Traverse forward with the 3rd operating state as the current state, and find the current state whose feature type is different from the previous state as the starting state; traverse backward with the 4th operating state as the current state, and find a current state whose feature type is different from the next state as the ending state. Then, calculate the control influence factor between the 3rd operating state and the 4th operating state based on the feature types and state times of the starting state and the ending state. If the control influence factor is positive, the 3rd operating state is the checkpoint where the operating state changes to stability.

[0088] Finally, input the multi-sensor data of the device under test into the signal synchronization model to obtain the feature representation of the device under test. For example, the multi-sensor data of the device under test includes a temperature value of 46 °C, a pressure value of 1.25 MPa, a current value of 10.3 A, a voltage value of 383 V, an operating speed value of 1460 r / min, a load percentage of 65%, etc. Input these data into the signal synchronization model, and the feature representation of the device is obtained as a 10-dimensional feature vector. Then, based on the trained synchronization control model, obtain the synchronization control result of the device under test. Five samples similar to the feature representation of the device under test can be found in the second training sample set. Based on the trained synchronization control model, obtain the synchronization control results of these five samples, and comprehensively evaluate the synchronization control results according to the control influence factors of the five samples to obtain the synchronization control result of the device under test. For example, the synchronization control results of these five samples are to reduce the load by 5%, reduce the load by 8%, reduce the load by 10%, reduce the load by 3%, and reduce the load by 6% respectively, and their control influence factors are 0.8, 0.6, 0.9, 0.7, and 0.85 respectively. Perform a weighted average on the synchronization control results according to these control influence factors to obtain the synchronization control result of the device under test as reducing the load by 6.8%. It is also possible to directly input the feature representation of the device under test into the trained synchronization control model, and the synchronization control model directly outputs the synchronization control result of the device under test, such as reducing the load by 7%, adjusting the motor speed to 1450 r / min, etc.

[0089] By applying the device automation testing and synchronization control method and system based on multi-sensor fusion, the automation testing efficiency of the automatic alignment switchgear in this factory has been increased by 30%, and the synchronization control accuracy has been increased by 40%. The synchronization performance of the device during operation is more stable, the failure rate has been reduced by 25%, and the service life of the device has been extended. At the same time, this method and system can quickly and accurately obtain the feature representation of the device and perform synchronization control, reducing the manual intervention cost and test time cost, improving the production efficiency and operation efficiency of the device, and saving a large amount of human and material resources for the enterprise.

[0090] The above embodiments of the present invention have the following beneficial effects: The present invention can improve the accuracy and control efficiency of power equipment automation testing. Through multi-sensor data fusion and signal synchronization model training, the operating state characteristics of the device can be accurately extracted and intelligently grouped, providing a reliable basis for subsequent control decisions. The synchronization control model constructed based on the feature type and control data can adaptively generate the optimal control strategy, and more accurate regulation can be implemented especially at the key nodes where the operating state changes to stability. The control influence factor calculation method can scientifically evaluate the control effect between adjacent states and ensure the rationality of sample selection. Through feature representation similarity matching and K-nearest neighbor comprehensive evaluation, the system response speed can be improved while ensuring control accuracy. The hierarchical design of the signal synchronization model can separately process the inherent parameters of the device and dynamic change characteristics to achieve a more comprehensive state representation. The present invention can form a closed-loop control system from state monitoring to intelligent decision-making and provide an intelligent solution for the operation and maintenance of power equipment.

[0091] As Figure 2 shown, a device automation test and synchronization control system based on multi-sensor fusion in some embodiments, the system includes: A signal synchronization model training module 201, configured to obtain multi-sensor data of an automatic point-switching switch cabinet and device operation state data, and construct a first training sample set; construct a signal synchronization model, and train the signal synchronization model based on the first training sample set to obtain a trained signal synchronization model; A feature grouping module 202, configured to perform feature grouping on the device operation state each time based on the trained signal synchronization model to obtain the feature type of each operation state; A synchronization control model training module 203, configured to construct a second training sample set based on the feature type of each operation state and the control data of the device between two operation states; construct a synchronization control model, and train the synchronization control model based on the second training sample set to obtain a trained synchronization control model; A synchronization control module 204, configured to input the multi-sensor data of the device to be tested into the signal synchronization model to obtain the feature representation of the device to be tested; and obtain the synchronization control result of the device to be tested based on the feature representation and the trained synchronization control model.

[0092] It can be understood that the various modules described in the device automation test and synchronization control system based on multi-sensor fusion correspond to the respective steps in the device automation test and synchronization control method described with reference to Figure 1 description. Therefore, the operations, features, and beneficial effects described above for the device automation test and synchronization control method based on multi-sensor fusion also apply to the device automation test and synchronization control system based on multi-sensor fusion and the modules included therein, and will not be elaborated herein.

[0093] In some embodiments, constructing a second training sample set based on the feature type of each operation state and the control data of the device between two operation states includes: Select the inspection of the transition of the operation state to stability, use the feature representation of this stability transition inspection as the input data of the sample, and use the control data of the device from this stability transition inspection to the next inspection as the label of the sample to construct a second training sample set.

[0094] It should be noted that the device automation test and synchronization control system based on multi-sensor fusion involved in the present invention is an integrated solution, aiming to achieve the automation test and precise synchronization control of the device through the collaborative work of multiple modules. The system analyzes and controls the operating state of the device by constructing a signal synchronization model and a synchronization control model, thereby improving the operating efficiency and stability of the device. This systematic solution can effectively integrate multi-sensor data and device operating state data, providing comprehensive technical support for the automation test and synchronization control of the device.

[0095] Specifically, the signal synchronization model training module is responsible for obtaining the multi-sensor data and device operating state data of the automatic point-switching switchgear and constructing the first training sample set. These data include the inherent parameters and dynamic operating parameters of the device. By using these data to train the signal synchronization model, it can accurately predict the signal synchronization situation of the device. The feature grouping module groups the features of the device's operating state each time based on the trained signal synchronization model, obtaining the feature types of each operating state. This process involves classifying the feature representations of the device's operating state so that the subsequent synchronization control model can perform targeted control according to different feature types. The synchronization control model training module is used to construct the second training sample set based on the feature types of each operating state and the control data of the device between two operating states, and train the synchronization control model. This module optimizes the performance of the synchronization control model by analyzing the control data of the device in different operating states. The synchronization control module inputs the multi-sensor data of the device to be tested into the signal synchronization model to obtain the feature representation of the device to be tested, and obtains the synchronization control result of the device to be tested according to the trained synchronization control model. This module is the core execution part of the system, directly determining the accuracy and efficiency of device synchronization control.

[0096] Preferably, during the implementation process, the operation steps of each module can be further refined. For example, in the signal synchronization model training module, data cleaning and feature engineering methods can be used to preprocess the obtained multi-sensor data and device operating state data to improve the training effect of the model. In the feature grouping module, clustering algorithms can be used to group the feature representations of the device's operating state, and the clustering results can be optimized through evaluation metrics. In the synchronization control model training module, cross-validation methods can be used to evaluate and optimize the model to ensure the generalization ability of the model. In the synchronization control module, a real-time feedback mechanism can be introduced to dynamically adjust the synchronization control strategy according to the actual operating conditions of the device. Through these refined operation steps, the performance and reliability of the system can be further improved, making it better adapt to complex industrial application scenarios.

[0097] Next, refer to Figure 3, which shows a schematic structural diagram of an electronic device 300 suitable for implementing some embodiments of the present invention. The electronic device in some embodiments of the present invention may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The terminal device shown is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present invention.

[0098] As Figure 3 shown, the electronic device 300 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 301, which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage device 308 into the random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other through a bus 304. The input / output (I / O) interface 305 is also connected to the bus 304.

[0099] Generally, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 308 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the electronic device 300 to communicate with other devices wirelessly or wirelessly to exchange data. Although Figure 3 the electronic device 300 with various devices is shown, it should be understood that it is not required to implement or include all the shown devices. More or fewer devices may be implemented or included alternatively. Figure 3 Each block shown in

[0100] Furthermore, the storage medium according to the embodiments of the present application stores program instructions capable of implementing all the above methods. Among them, the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0101] Embodiment 2 In the application of the automatic alignment switchgear, by acquiring multi-sensor data and operating status data of power equipment, and using the signal synchronization model and the synchronization control model, the automatic testing and precise synchronization control of the equipment are realized.

[0102] Real-time data is collected from various sensors (such as temperature sensors, pressure sensors, current sensors, etc.) installed on the automatic alignment switchgear. The operating parameters of the equipment (such as voltage, current, power, etc.) and the execution status of control instructions are recorded. The collected multi-sensor data and equipment operating status data are integrated to construct a first training sample set containing various equipment statuses. The first training sample set is trained using machine learning algorithms (such as neural networks) to obtain a signal synchronization model capable of predicting the signal synchronization situation of the equipment.

[0103] Based on the trained signal synchronization model, feature extraction and grouping are performed on the operating status of the equipment each time to obtain the feature types of each operating status. Checkpoint data when the operating status changes to stable is selected, and combined with the control data between two states, a second training sample set is constructed. The second training sample set is used to train the synchronization control model so that it can output a synchronization control result according to the input feature representation. The multi-sensor data of the equipment to be tested is input into the trained signal synchronization model to obtain the feature representation of the equipment. Based on the feature representation, the synchronization control result of the equipment is output using the synchronization control model. According to the synchronization control result, corresponding control instructions are sent to the equipment to realize the automatic testing and synchronization control of the equipment.

[0104] Through automatic testing and synchronization control, manual intervention is reduced and the testing efficiency is improved. By using multi-sensor data fusion and machine learning models, the precise perception and control of the equipment status are realized, and the control accuracy is improved. Through real-time monitoring and synchronization control, abnormal equipment statuses are detected and corrected in a timely manner to ensure the stable operation of the equipment.

[0105] The above description is only some preferred embodiments of the present invention and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the embodiments of the present invention.

Claims

1. An automated testing and synchronous control method for a device based on multi-sensor fusion, characterized in that, It includes the following steps: Obtain the multi-sensor data and device operation status data of the automatic point-switching switchgear, and construct the first training sample set; Construct a signal synchronization model, and train the signal synchronization model based on the first training sample set to obtain a trained signal synchronization model; Based on the trained signal synchronization model, perform feature grouping on each device operation status to obtain the feature types of each operation status; Construct a second training sample set based on the feature types of each operation status and the control data of the device between two operation statuses; Construct a synchronization control model, and train the synchronization control model based on the second training sample set to obtain a trained synchronization control model; Input the multi-sensor data of the device to be tested into the signal synchronization model to obtain the feature representation of the device to be tested; based on the feature representation, obtain the synchronization control result of the device to be tested based on the trained synchronization control model.

2. The automated testing and synchronous control method for a device based on multi-sensor fusion according to claim 1, characterized in that, Constructing a second training sample set based on the feature types of each operation status and the control data of the device between two operation statuses includes: Select the inspection of the transition from the operating state to the stable state, use the feature representation of this stable transition inspection as the input data of the sample, and use the control data of the device between this stable transition inspection and the next inspection as the label of the sample to construct the second training sample set.

3. The automated testing and synchronous control method for a device based on multi-sensor fusion according to claim 2, characterized in that, Select the inspection of the transition from the operating state to the stable state in the following way: For each device, calculate the control influence factor between two adjacent operation statuses; If the control influence factor between the t-th operation status and the (t + 1)-th operation status is positive, then the t-th operation status is the inspection of the transition from the operating state to the stable state.

4. The automated testing and synchronous control method for a device based on multi-sensor fusion according to claim 3, characterized in that, Calculate the control influence factor between the t-th operation status and the (t + 1)-th operation status in the following way: Traverse forward in turn with the t-th operation status as the current state until a current state with a different feature type from a previous state is found as the starting state; Traverse backward in turn with the (t + 1)-th operation status as the current state until a current state with a different feature type from a subsequent state is found as the ending state; Calculate the control influence factor between the t-th operation status and the (t + 1)-th operation status based on the feature types and state times of the starting state and the ending state.

5. The automated testing and synchronous control method for a device based on multi-sensor fusion according to claim 4, characterized in that, Calculate the control influence factor between the t-th operation status and the (t + 1)-th operation status based on the feature types and state times of the starting state and the ending state using the following formula: Among them, the feature type of the starting state is the i-th feature type, and the feature type of the ending state is the j-th feature type. represents the distance between the feature type of the starting state and the feature type of the ending state. represents the time of the starting state. represents the time of the ending state. represents the hyperbolic tangent function. represents the sign function. and represents a parameter. represents the control influence factor between the t-th running state and the (t + 1)-th running state.

6. The automated testing and synchronous control method for a device based on multi-sensor fusion according to claim 1, characterized in that, Obtaining the synchronization control result of the device to be tested based on the feature representation and the trained synchronization control model includes: Search for K samples similar to the feature representation of the device to be tested in the second training sample set; Obtain the synchronization control results of the K samples based on the trained synchronization control model; Comprehensively evaluate the synchronization control results of the K samples based on the control influence factors of the K samples to obtain the synchronization control result of the device to be tested.

7. The automated testing and synchronous control method for a device based on multi-sensor fusion according to claim 1, characterized in that, Performing feature grouping on each device operation status based on the trained signal synchronization model to obtain the feature types of each operation status includes: For the t-th operation status of the device, input the t-th operation status and the previous operation status data into the trained signal synchronization model to obtain the feature representation of the t-th operation status of the device and the signal synchronization prediction result; Feature grouping is performed on the feature representations of all operating states of all devices to obtain multiple feature types; For each feature type, calculate the average value of the signal synchronization prediction results corresponding to the feature representations belonging to this feature type, and sort the feature types according to the average value of the signal synchronization prediction results to obtain the serial number of each feature type; The feature type of each operating state is the feature type to which the feature representation of this operating state belongs.

8. The method for automated testing and synchronous control of a device based on multi-sensor fusion according to claim 1, wherein, The signal synchronization model includes: An equipment inherent parameter extraction module for extracting inherent attribute features from the equipment inherent parameters of the samples; A dynamic change feature extraction module for extracting dynamic change features from the operating state data of the samples; A feature fusion module for fusing the inherent attribute features and the dynamic change features; A prediction module for performing signal synchronization prediction based on the fused features.

9. An automated testing and synchronous control system for a device based on multi-sensor fusion, wherein, It includes: A signal synchronization model training module for obtaining the multi-sensor data and equipment operating state data of the automatic point-switching switchgear, and constructing a first training sample set; Constructing a signal synchronization model, and training the signal synchronization model based on the first training sample set to obtain a trained signal synchronization model; A feature grouping module for performing feature grouping on each operating state of the equipment based on the trained signal synchronization model to obtain the feature type of each operating state; A synchronization control model training module for constructing a second training sample set based on the feature type of each operating state and the control data of the equipment between two operating states; Constructing a synchronization control model, and training the synchronization control model based on the second training sample set to obtain a trained synchronization control model; A synchronization control module for inputting the multi-sensor data of the device to be tested into the signal synchronization model to obtain the feature representation of the device to be tested; and obtaining the synchronization control result of the device to be tested based on the feature representation and the trained synchronization control model.

10. The automated testing and synchronous control system for a device based on multi-sensor fusion according to claim 9, wherein, Constructing a second training sample set based on the feature type of each operating state and the control data of the equipment between two operating states includes: Selecting the inspection of the transition of the operating state to stability, using the feature representation of this stability transition inspection as the input data of the sample, and using the control data of the equipment between this stability transition inspection and the next inspection as the label of the sample to construct a second training sample set.

Images