Inductor powder forming part EMC defect diagnosis system based on multi-mode perception

Abstract

The invention discloses an inductance powder forming part EMC defect diagnosis system based on multi-mode perception, and relates to the technical field of inductance powder forming part testing, and the system comprises a model construction module which is used for formulating a plurality of pieces of test information for an inductance powder forming part, and constructing corresponding test models based on the plurality of pieces of test information; wherein the test information comprises environment information and control information; and the integration module is used for determining a test scheme of the inductance powder forming part and integrating the plurality of test models based on the test scheme to obtain an integrated model. According to the method, a certain time window can be provided for transfer of the test data, the test data can be prevented from being stolen and tampered by an unauthorized port, the safety and accuracy of the test data in transmission can be ensured, accurate test data can be provided for a subsequent diagnosis model, and the test efficiency is improved. And the diagnosis model can conveniently output an accurate diagnosis result according to the accurate test data.

Description

Technical Field

[0001] This invention relates to the field of testing technology for inductor powder molded parts, and more specifically to an EMC defect diagnosis system for inductor powder molded parts based on multimodal sensing. Background Technology

[0002] Inductor powder molded parts, as core passive components in electronic devices, are widely used in key areas such as new energy vehicle charging piles, industrial frequency converters, and home appliance controllers due to their high magnetic permeability, low loss, and compact structure. The electromagnetic compatibility (EMC) performance of inductor powder molded parts directly determines the overall performance and compliance of related equipment. Therefore, accurate and efficient EMC defect diagnosis of inductor powder molded parts is a crucial step in ensuring the reliability and stability of electronic equipment.

[0003] During the testing of inductor powder molded parts, test data such as electrical signals, magnetic signals, and environmental parameters need to be collected. However, existing methods often use a single transmission channel or shared transmission resources, which can easily lead to data confusion. At the same time, test data (including sensitive / performance parameters of the product) lacks an effective security protection mechanism during transmission, and the test data is at risk of being stolen or tampered with. Summary of the Invention

[0004] The purpose of this invention is to provide an EMC defect diagnosis system for inductive powder molded parts based on multimodal sensing, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an EMC defect diagnosis system for inductive powder molded parts based on multimodal sensing, comprising: The model building module is used to define multiple test information for inductor powder molded parts and build corresponding test models based on these test information; the test information includes environmental information and control information. An integration module is used to determine the test plan for inductor powder molded parts. Based on the test plan, multiple test models are integrated to obtain an integrated model. The configuration module is used to configure the cloud server for the integrated model, set multiple collection points and multiple receiving points for the integrated model and the cloud server respectively, and establish a one-to-one transmission channel between the multiple collection points and multiple receiving points. The test execution module is used to test the inductor powder molded parts through the integrated model and transmit the test data collected at the acquisition points to the receiving points through the transmission channel; The diagnostic module is used to preprocess the test data received from multiple receiving points, and input the preprocessed test data into the trained diagnostic model. The trained diagnostic model outputs the diagnostic results of the test data.

[0006] In a preferred embodiment, the step of defining multiple test information for the inductor powder molded part and constructing corresponding test models based on the multiple test information includes: Determine the product information of the inductor powder molded part, and formulate the corresponding test information for the inductor powder molded part based on the product information. The test information includes environmental information and control information. A test model is constructed based on the test information corresponding to the inductor powder molded part, and the test information is used as the label information of the corresponding test model.

[0007] In a preferred embodiment, the step of determining the test plan for the inductor powder molded part involves integrating multiple test models based on the test plan to obtain an integrated model, including: A test plan for inductor powder molded parts is developed based on the product information of the inductor powder molded parts; the test plan includes: the execution order of multiple test models and the sampling frequency of multiple test models; Multiple test models are arranged according to the test plan, and the arranged test models are combined to form an integrated model.

[0008] In a preferred embodiment, configuring a cloud server for the integrated model, setting multiple acquisition points and multiple receiving points for both the integrated model and the cloud server, and establishing a one-to-one transmission channel between the multiple acquisition points and the multiple receiving points includes: Configure a cloud server for multiple test models in the integrated model; Multiple data collection points are deployed for each of the multiple test models in the integrated model. These data collection points are used to collect test data for the corresponding test models. For each collection point, multiple receiving points are deployed within the cloud server. The receiving points are used to receive the test data collected by the collection points through the transmission channel. The number of collection points and receiving points is the same, and a transmission channel is established between the corresponding collection points and receiving points. A distribution center is deployed within the cloud server. The distribution center deploys transmission carriers for corresponding transmission channels based on the collection points. Each transmission carrier is configured with a connection port and given a unique number. Each transmission carrier is spatially divided into a primary data space and multiple secondary data spaces. External storage points and internal storage points are configured for both the primary and secondary data spaces. The internal storage points are located within the external storage points. A first connection point and a second connection point are configured for both the external and internal storage points. The internal storage points are used to store test data collected by the collection points, while the external storage points are used to store spoofed data. A virtual surface is constructed within the cloud server, and multiple acquisition points, receiving points, transmission channels, and transmission carriers are mapped onto the virtual surface to obtain the positions of multiple acquisition points, receiving points, transmission channels, and transmission carriers on the virtual surface; each position is connected one-to-one with the corresponding acquisition point / receiving point / transmission channel / transmission carrier.

[0009] In a preferred embodiment, in the virtual surface: Multiple transmission carriers corresponding to the same acquisition cycle are divided into an adjacent group, and multiple transmission carriers within the adjacent group are topologically connected based on their physical distance. The number of transmission carriers between two transmission carriers in the adjacent group is used as the distance coefficient between the two transmission carriers. The larger the distance coefficient, the farther away the two transmission carriers are, and the smaller the distance coefficient, the closer the two transmission carriers are. Trigger conditions are set for the transmission carrier. The virtual plane monitors the transmission carrier through the trigger conditions and marks the transmission carrier that meets the trigger conditions as an abnormal transmission carrier. The trigger condition is that the transmission carrier is accessed by an unauthorized port.

[0010] In a preferred embodiment, the step of testing the inductor powder molded part using an integrated model and transmitting the test data collected from multiple sampling points through a transmission channel includes: Multiple test models within the integrated model are activated to test the inductor powder molded parts. The acquisition points corresponding to each test model collect the test data of the inductor powder molded parts through the acquisition cycle. The data collection point stores the collected test data into the corresponding transmission carrier, which then transmits the data to the corresponding receiving point through the transmission channel. During the transmission process, the virtual surface monitors the transmission carrier in real time.

[0011] In a preferred embodiment, the virtual surface performs real-time monitoring of the transmission medium during transmission, including: When the virtual plane detects that the connection port of the transmission carrier is accessed by an unauthorized port, the transmission carrier is marked as an abnormal transmission carrier. The abnormal transmission carrier will connect the first docking point in the auxiliary data space to the unauthorized port. Meanwhile, the transmission carrier with the smallest distance coefficient between the adjacent group and the abnormal transmission carrier is selected as the backup transmission carrier. The distribution center sends the same password token to the second docking point of the abnormal transmission carrier and the backup transmission carrier through the point. The abnormal transmission carrier and the backup transmission carrier establish an encrypted transmission channel through the preset connection rules. The encrypted transmission channel transmits the abnormal transmission carrier's number and test data to the internal storage point of the backup transmission carrier's auxiliary data space. The connection rules include establishing an encrypted communication channel between two transmission carriers with the same password token on the virtual plane; When the faulty transmission carrier and the backup transmission carrier arrive at their respective receiving points, the receiving points will send the test data marked with the corresponding number from the backup transmission carrier to the receiving point corresponding to the faulty transmission carrier.

[0012] In a preferred embodiment, the preprocessing of test data received from multiple receiving points, and the input of the preprocessed test data into a trained diagnostic model, wherein the trained diagnostic model outputs diagnostic results of the test data, includes: Collect and preprocess multiple historical test data and their corresponding actual diagnostic results, construct a historical dataset from the multiple historical test data and their corresponding actual diagnostic results, and use the historical dataset to train the diagnostic model to obtain a trained diagnostic model. The preprocessed test data is input into the trained diagnostic model, which then outputs the diagnostic results corresponding to the test data.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention, by connecting a first docking point to an unauthorized port when the connection port of the transmission carrier is accessed by an unauthorized port, serves to confuse the unauthorized port's access. This facilitates the abnormal transmission carrier establishing an encrypted communication channel with the second docking point of the backup transmission carrier through a second docking point, and transferring the test data within the abnormal transmission carrier. This provides a certain time window for the transfer of test data, thereby preventing test data from being stolen or tampered with by unauthorized ports, ensuring the security and accuracy of test data during transmission, and providing accurate test data for subsequent diagnostic models. This allows the diagnostic models to output accurate diagnostic results based on accurate test data. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a system block diagram of the present invention.

[0016] Figure 2 This is a schematic diagram of the transmission channel structure in this invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 and Figure 2 As shown in this embodiment, the EMC defect diagnosis system for inductive powder molded parts based on multimodal sensing includes: The model building module is used to define multiple test information for inductor powder molded parts and build corresponding test models based on these test information; the test information includes environmental information and control information. An integration module is used to determine the test plan for inductor powder molded parts. Based on the test plan, multiple test models are integrated to obtain an integrated model. The configuration module is used to configure the cloud server for the integrated model, set multiple collection points and multiple receiving points for the integrated model and the cloud server respectively, and establish a one-to-one transmission channel between the multiple collection points and multiple receiving points. The test execution module is used to test the inductor powder molded parts through the integrated model and transmit the test data collected at the acquisition points to the receiving points through the transmission channel; The diagnostic module is used to preprocess the test data received from multiple receiving points, and input the preprocessed test data into the trained diagnostic model. The trained diagnostic model outputs the diagnostic results of the test data.

[0019] In this embodiment, when a transmission carrier is accessed by an unauthorized port, the virtual surface immediately marks the transmission carrier as an abnormal transmission carrier. The abnormal transmission carrier temporarily connects its first docking point in its secondary data space to the unauthorized port. The external storage point of the secondary data space stores spoofed data, preventing the unauthorized port from directly accessing the test data in the main data space, thus misleading attackers and isolating the test data. Simultaneously, based on the transmission carrier's grouping and proximity coefficient, the virtual surface prioritizes the transmission carrier with the smallest proximity coefficient from the adjacent group of the abnormal transmission carrier as a backup transmission carrier. The distribution center within the cloud server, through the corresponding point on the virtual surface, sends data to the abnormal transmission carrier and... The second docking point of the backup transmission carrier sends the same password token; according to the preset connection rules, two transmission carriers holding the same password token on the virtual surface can pass the consistency verification and then establish an encrypted communication channel. The abnormal transmission carrier transmits the test data stored in its internal storage point and its own unique number completely to the internal storage point of the backup transmission carrier's auxiliary data space through the encrypted communication channel. This realizes the secure transfer of test data in the abnormal transmission carrier, thereby preventing test data from being stolen or tampered with by unauthorized ports, ensuring the security and accuracy of test data during transmission, and providing accurate test data for subsequent diagnostic models, so that the diagnostic models can output accurate diagnostic results based on accurate test data.

[0020] In one embodiment, the step of defining multiple test information for the inductor powder molded part and constructing corresponding test models based on the multiple test information includes: Determine the product information of the inductor powder molded part, and formulate the corresponding test information for the inductor powder molded part based on the product information. The test information includes environmental information and control information. A test model is constructed based on the test information corresponding to the inductor powder molded part, and the test information is used as the label information of the corresponding test model.

[0021] It should be noted that the product information for powder-molded inductors specifically includes the product model (e.g., EE series, PQ series), core material (e.g., iron powder core, ferrite, nanocrystalline alloy), rated electrical parameters (rated current, operating frequency, rated voltage, inductance), structural parameters (size, number of winding turns, packaging method), application scenarios (e.g., new energy vehicle charging piles, home appliance controllers, industrial frequency converters), and the EMC (electromagnetic compatibility) standards that must be met. Based on the product information, environmental and control information for the powder-molded inductors should be developed. Environmental information includes simulated operating conditions such as temperature, humidity, and air pressure. The control information includes electrical and magnetic test signals designed for EMC defect diagnosis. Specifically, the electrical test signals are set according to the rated operating frequency of the inductor powder molded part and the EMC test requirements, selecting appropriate waveforms (sine wave, square wave, pulse wave), frequency ranges (from 50Hz power frequency to MHz radio frequency), and voltage / current amplitudes (not exceeding 1.2 times the rated parameters of the inductor to avoid damaging the product). The magnetic test signals are applied through an external magnetic field generator, and the magnetic field strength needs to be referenced to the magnetic field interference level of the product's environment (e.g., 0.5mT~5mT in industrial environments, 0.1mT~1mT in household environments).

[0022] A test model is established based on test information, and the test information is bound to the test model. A hardware test platform is built, including inductor clamping fixtures, signal generators, magnetic field generators, data acquisition devices, etc. The hardware devices in the hardware test platform can accurately reproduce environmental and control information. The test model is an executable test configuration or test unit that controls the corresponding test information in the hardware test platform. By formulating test information based on the product information of the inductor powder molded parts and building a test model based on the test information, it is easy to quickly retrieve the corresponding test model for subsequent testing of the inductor powder molded parts.

[0023] In one embodiment, determining the test plan for the inductor powder molded part involves integrating multiple test models based on the test plan to obtain an integrated model, including: A test plan for inductor powder molded parts is developed based on the product information of the inductor powder molded parts; the test plan includes: the execution order of multiple test models and the sampling frequency of multiple test models; Multiple test models are arranged according to the test plan, and the arranged test models are combined to form an integrated model.

[0024] It should be noted that the test plan is developed based on the product information of the inductor powder molded parts. The test plan includes the execution sequence and sampling frequency of the test models. The execution sequence is designed according to the principle of "from simple to complex, from no interference to interference, and from normal operating conditions to extreme operating conditions". This can avoid the cross-influence of interference factors from different test models and ensure the accuracy of the basic test data. For example, the room temperature test model without external magnetic field interference is executed first to obtain the baseline electromagnetic response data of the inductor powder molded parts under standard conditions; then the room temperature test model with magnetic field interference is executed to compare and analyze the impact of magnetic field on inductor performance; finally, the composite operating condition test model of high temperature and high humidity + magnetic field interference is executed to verify the EMC stability of the product in extreme environments. If the order is reversed, and the composite operating condition test is performed first, the inductor performance may change irreversibly due to extreme conditions, affecting the accuracy of subsequent routine operating condition tests. The sampling frequency setting is based on the characteristics of the test signal in the control information and the Nyquist sampling theorem to ensure that the collected test data can completely restore the signal waveform. Generally, the sampling frequency should be at least twice the highest frequency of the test signal. In practical applications, to improve data accuracy, it is usually 5 to 10 times. For example, if the highest frequency of the electrical test signal is 10MHz, the sampling frequency should be set to 50MHz to 100MHz. According to the execution order in the test plan, multiple test models are arranged sequentially on the software platform (such as industrial control software). The arranged test models form an integrated model. Thus, the integrated model can call each test model in the order of arrangement, and complete the entire process test without manual intervention. This facilitates the collection of test data under the corresponding test information by each test model, which provides complete data support for subsequent EMC defect diagnosis.

[0025] In one embodiment, configuring a cloud server for the integrated model, setting multiple collection points and multiple receiving points for both the integrated model and the cloud server, and establishing a one-to-one transmission channel between the multiple collection points and the multiple receiving points includes: Configure a cloud server for multiple test models in the integrated model; Multiple data collection points are deployed for each of the multiple test models in the integrated model. These data collection points are used to collect test data for the corresponding test models. For each collection point, multiple receiving points are deployed within the cloud server. The receiving points are used to receive the test data collected by the collection points through the transmission channel. The number of collection points and receiving points is the same, and a transmission channel is established between the corresponding collection points and receiving points. A distribution center is deployed within the cloud server. The distribution center deploys transmission carriers for corresponding transmission channels based on the collection points. Each transmission carrier is configured with a connection port and given a unique number. Each transmission carrier is spatially divided into a primary data space and multiple secondary data spaces. External storage points and internal storage points are configured for both the primary and secondary data spaces. The internal storage points are located within the external storage points. A first connection point and a second connection point are configured for both the external and internal storage points. The internal storage points are used to store test data collected by the collection points, while the external storage points are used to store spoofed data. A virtual surface is constructed within the cloud server, and multiple acquisition points, receiving points, transmission channels, and transmission carriers are mapped onto the virtual surface to obtain the positions of multiple acquisition points, receiving points, transmission channels, and transmission carriers on the virtual surface; each position is connected one-to-one with the corresponding acquisition point / receiving point / transmission channel / transmission carrier.

[0026] It should be noted that cloud servers are configured for the integrated model, and multiple acquisition points are deployed for each test model within the integrated model. For example, if each test model needs to simultaneously acquire multiple types of data such as electrical signal amplitude, phase, magnetic field strength, and frequency, then multiple acquisition modules (such as voltage acquisition modules and magnetic field acquisition modules) need to be deployed accordingly. Each acquisition module is treated as an independent acquisition point, and the acquisition points are directly deployed on the hardware test platform corresponding to the test model to ensure accurate acquisition of test data generated during the test according to the (acquisition cycle). Corresponding to the number of acquisition points, an equal number of receiving points are deployed within the cloud server. The receiving points are essentially virtual receiving processes running within the cloud server, and each receiving point is bound to a unique network port specifically for listening to data transmission requests from the corresponding acquisition point. Since there is a one-to-one correspondence between acquisition points and receiving points, a transmission channel is established between each set of corresponding (matched) acquisition points and receiving points. This one-to-one channel design can prevent confusion between test data acquired by different acquisition points.

[0027] A distribution center is deployed within the cloud server. This distribution center is a scheduler within the cloud server. It can allocate independent transmission carriers to each transmission channel through collection points. The transmission carriers are virtual machines, generated by the collection points according to the collection cycle. Each transmission carrier is configured with a unique connection port and number, which must be associated with the IDs of the corresponding collection and receiving points (e.g., collection point C001, receiving point R001 corresponding to transmission carrier T001), facilitating rapid location of the test data flow path. Each transmission carrier is spatially divided, with one main data space and multiple auxiliary data spaces (typically 2-3). The main data space stores core test data, while the auxiliary data spaces store test data from other transmission carriers and connect the first docking point to the unauthorized port when the transmission carrier meets trigger conditions. Both the main and auxiliary data spaces have internal and external storage points. The internal storage points store the test data collected by the collection points; the external storage points store dummy data generated based on the data from the internal storage points. The dummy data is generated through random perturbation or interpolation to mimic the format, numerical range, and data structure of the test data (e.g., test data is 2). The voltage signal is 5V±0.2V, while spurious data generates random voltage values ​​of 2.4V-2.6V (with sampling frequency and data length consistent with the test data). Spurious data is generally pre-generated based on historical test data collected at the acquisition point when the transmission carrier is generated, and stored in the external storage point of the transmission carrier. Both the main and auxiliary data spaces have external storage points storing spurious data. This can mislead unauthorized ports when the connection port of the transmission carrier is accessed by unauthorized ports, thus protecting the test data in the internal storage point. A first docking point is configured for the external storage point, and a second docking point is configured for the internal storage point. The internal and external storage points are essentially virtual machines. The first and second docking points are virtual communication ports used to connect the transmission carrier with other transmission carriers. When the connection port of the transmission carrier is accessed by unauthorized ports, the first docking point docks with the unauthorized port, thereby confusing the unauthorized port access. It also facilitates the abnormal transmission carrier to establish an encrypted communication channel with the second docking point of the backup transmission carrier through the second docking point, and to transfer the test data in the abnormal transmission carrier, providing a certain time window for the transfer of test data. By constructing a virtual surface within the cloud server, which is a virtual management interface built based on visualization technologies (such as Unity3D and WebGL), all components (collection points / receiving points / transmission channels / transmission carriers) are mapped to the interface according to their actual topology. Each component corresponds to a unique point on the virtual surface, and the identifier of the point is consistent with the component's ID (e.g., transmission carrier T001 corresponds to point P001 on the virtual surface). The virtual surface can display the abstract transmission channels, and maintenance personnel can directly view the operating status of each component and test the data transmission progress through the virtual surface, which facilitates the grouping of transmission carriers and anomaly monitoring.

[0028] In one embodiment, in the virtual surface: Multiple transmission carriers corresponding to the same acquisition cycle are divided into an adjacent group, and multiple transmission carriers within the adjacent group are topologically connected based on their physical distance. The number of transmission carriers between two transmission carriers in the adjacent group is used as the distance coefficient between the two transmission carriers. The larger the distance coefficient, the farther away the two transmission carriers are, and the smaller the distance coefficient, the closer the two transmission carriers are. Trigger conditions are set for the transmission carrier. The virtual plane monitors the transmission carrier through the trigger conditions and marks the transmission carrier that meets the trigger conditions as an abnormal transmission carrier. The trigger condition is that the transmission carrier is accessed by an unauthorized port.

[0029] It should be noted that by dividing transmission carriers with the same acquisition cycle into adjacent groups, for example, among 6 transmission carriers, T001, T002, and T003 correspond to the same acquisition cycle and are deployed in the same subnet, they are divided into adjacent group 1; then, the transmission carriers within each group are topologically connected to form a mesh or chain topology, and a proximity coefficient is defined for the transmission carriers within the group: the number of transmission carriers between two transmission carriers is used as the metric; the more intervals, the larger the proximity coefficient, and the farther apart the two are; the fewer intervals, the smaller the proximity coefficient, and the farther apart the two are. The closer they are; for example, in group 1, there is no gap between T001 and T002, so the distance coefficient is 0, meaning they are the closest; T001 and T003 are separated by T002, so the distance coefficient is 1, meaning they are the next closest. The distance coefficient can provide a priority basis when selecting a backup transmission carrier in the event of an abnormal transmission carrier. Thus, by setting adjacent groups and distance coefficients, when selecting a backup transmission carrier for an abnormal transmission carrier, the transmission carrier on the transmission channel with the closer physical distance can be selected, reducing the transmission time for the abnormal transmission carrier to transfer test data to the backup transmission carrier. The system triggers access to a transmission carrier via an unauthorized port. An unauthorized port is a network port not registered in the system used for data reception or access (such as a port used for external malicious attacks or an unauthorized maintenance port). The system monitors the connection port access status of each transmission carrier in real time via a virtual plane. For example, by comparing port access logs with the system's preset list of authorized ports, once an unauthorized port attempts to access a transmission carrier, the carrier is immediately marked as an abnormal transmission carrier. The distribution center sends the same password token to the second connection point of both the abnormal and backup transmission carriers. The abnormal and backup transmission carriers establish an encrypted transmission channel through preset connection rules. This encrypted transmission channel transmits the abnormal transmission carrier's ID and test data to the internal storage point of the backup transmission carrier's secondary data space. This ensures that abnormal transmission carriers are detected promptly, facilitating subsequent transfer of test data and effectively preventing security risks such as theft and tampering of test data, thus providing security for the transmission of test data.

[0030] In one embodiment, the testing of the inductor powder molded part using an integrated model and the transmission of test data collected from multiple sampling points through a transmission channel includes: Multiple test models within the integrated model are activated to test the inductor powder molded parts. The acquisition points corresponding to each test model collect the test data of the inductor powder molded parts through the acquisition cycle. The data collection point stores the collected test data into the corresponding transmission carrier, which then transmits the data to the corresponding receiving point through the transmission channel. During the transmission process, the virtual surface monitors the transmission carrier in real time.

[0031] In one embodiment, the virtual surface performs real-time monitoring of the transmission medium during transmission, including: When the virtual plane detects that the connection port of the transmission carrier is accessed by an unauthorized port, the transmission carrier is marked as an abnormal transmission carrier. The abnormal transmission carrier will connect the first docking point in the auxiliary data space to the unauthorized port. Meanwhile, the transmission carrier with the smallest distance coefficient between the adjacent group and the abnormal transmission carrier is selected as the backup transmission carrier. The distribution center sends the same password token to the second docking point of the abnormal transmission carrier and the backup transmission carrier through the point. The abnormal transmission carrier and the backup transmission carrier establish an encrypted transmission channel through the preset connection rules. The encrypted transmission channel transmits the abnormal transmission carrier's number and test data to the internal storage point of the backup transmission carrier's auxiliary data space. The connection rules include establishing an encrypted communication channel between two transmission carriers with the same password token on the virtual plane; When the abnormal transmission carrier and the backup transmission carrier arrive at the corresponding receiving point, the receiving point will send the test data marked with the corresponding number in the backup transmission carrier to the receiving point corresponding to the abnormal transmission carrier. It should be noted that multiple test models within the integrated model are activated, and each test model sequentially tests the inductor powder molded part. Each test model's corresponding acquisition point collects test data from the inductor powder molded part during the testing process through a collection cycle. Immediately after collection, the test data is uploaded to the transmission carrier corresponding to the acquisition point. The test data is stored in the inner storage point of the main data space within the transmission carrier. The outer storage point stores spoofed data, and the inner storage point is located inside the outer storage point. This allows the outer storage point and its stored spoofed data to confuse unauthorized port access, thus providing security protection for the test data. The transmission carrier transmits the stored test data to the corresponding receiving point within the cloud server through a transmission channel. Throughout the transmission process, the virtual plane continuously monitors the transmission carrier's operating status, port access status, and data transmission rate in real time, ensuring the continuity and security of data transmission. If unauthorized port access is detected on the transmission carrier, the first docking point corresponding to the auxiliary data space within the transmission carrier is immediately docked with the unauthorized port.

[0032] When a connection port of a transmission carrier is detected to be accessed by an unauthorized port not registered in the system, the virtual plane immediately marks the transmission carrier as an abnormal transmission carrier and sends the corresponding number to the corresponding receiving point. Upon receiving the abnormal transmission carrier, the receiving point destroys it according to the number. At this time, the abnormal transmission carrier temporarily connects its first docking point in its secondary data space to the unauthorized port. The external storage point of the secondary data space stores dummy data, which can prevent the unauthorized port from directly accessing the test data in the main data space, thus confusing attackers and isolating the test data. Simultaneously, based on the grouping and proximity coefficient of the transmission carrier, the virtual plane prioritizes selecting the transmission carrier with the smallest proximity coefficient (closest distance) from the adjacent group of the abnormal transmission carrier. As a backup transmission carrier, the distribution center within the cloud server sends the same password token (the password token is dynamically generated once to ensure transmission security) to the second connection point of both the abnormal transmission carrier and the backup transmission carrier through the corresponding point on the virtual surface. According to preset connection rules, the two transmission carriers holding the same password token on the virtual surface can pass the consistency verification and establish an encrypted communication channel. This encrypted communication channel is used to transmit test data to prevent the test data from being stolen or tampered with during the transfer process. Subsequently, the abnormal transmission carrier transmits the test data stored in its internal storage point and its own unique number (to facilitate the receiving point to identify the ownership of the test data) completely to the internal storage point of the backup transmission carrier's auxiliary data space through the encrypted communication channel, thereby realizing the secure transfer of test data within the abnormal transmission carrier.

[0033] Once both the faulty transmission carrier and the backup transmission carrier arrive at their respective receiving points, the receiving points destroy the faulty transmission carrier. Simultaneously, the receiving point of the backup transmission carrier sends the test data marked with the faulty transmission carrier number within the backup transmission carrier to the receiving point corresponding to the original faulty transmission carrier. This ensures that the ownership of the test data is not confused, providing accurate test data for subsequent diagnostic models, and facilitating the diagnostic models to output accurate diagnostic results based on accurate test data.

[0034] In one embodiment, the preprocessing of test data received from multiple receiving points, and the input of the preprocessed test data into a trained diagnostic model, wherein the trained diagnostic model outputs the diagnostic results of the test data, includes: Collect and preprocess multiple historical test data and their corresponding actual diagnostic results, construct a historical dataset from the multiple historical test data and their corresponding actual diagnostic results, and use the historical dataset to train the diagnostic model to obtain a trained diagnostic model. The preprocessed test data is input into the trained diagnostic model, which then outputs the diagnostic results corresponding to the test data.

[0035] It should be noted that historical test data of inductive powder molded parts under different batches and working conditions were collected (including electrical signal data, magnetic signal data, temperature value, humidity value, and air pressure under various test information such as no defects, excessive electromagnetic leakage, and insufficient anti-magnetic interference). At the same time, the actual diagnostic results corresponding to each data point were recorded (such as "no defects", "excessive electromagnetic leakage (amplitude deviation > 0.3V)", "insufficient immunity (signal fluctuation > 10% under magnetic field interference)", "normal"). The historical test data is preprocessed in three steps: First, data cleaning is performed to remove outliers (such as extreme data exceeding the normal range by 10 times) caused by equipment failure or transmission interference, and to fill in missing values ​​(using the mean of adjacent data or interpolation). Second, data standardization / normalization is performed to convert data of different dimensions and magnitudes (such as voltage signals of 2-3V and magnetic field strength of 0.5-1mT) into a unified range (such as 0-1 or -1-1) to eliminate the impact of magnitude differences on model training. Third, feature extraction is performed to extract key features that characterize EMC defects from the test data (such as signal peak value, fluctuation variance, frequency response peak value, magnetic field coupling coefficient, etc.) to reduce data dimensionality while retaining core information.Preprocessed historical test data was paired with corresponding diagnostic results. Historical test data from the same batch and under the same working conditions, along with their corresponding actual diagnostic results, were considered as one sample. Multiple samples constituted a complete historical dataset, which was divided into training and validation sets in a 7:3 ratio. A CNN model was selected to build the diagnostic model. The model was trained using 700 training sets and evaluated using 300 validation sets. The final diagnostic model achieved an accuracy of 95%, and the trained model was saved. Specifically, the diagnostic model was built using convolutional layers, pooling layers, and fully connected layers. Convolutional layers were used to extract local correlation information of key features from the test data. Pooling layers achieved feature dimensionality reduction to improve model computational efficiency. Fully connected layers mapped the extracted high-dimensional features to a preset defect category space, adapting to the rapid diagnostic needs in industrial scenarios. During training, the preprocessed 6-dimensional key feature data (including signal peak value, fluctuation variance, etc.) was first divided into 700 training sets in a 7:3 ratio. The model was trained on a training set and a validation set of 300 sets. The difference between the model's predicted values ​​and the actual diagnostic results was measured using the cross-entropy loss function. The convolutional kernel parameters and fully connected layer weights were iteratively adjusted using the gradient descent algorithm. At the same time, the model's generalization ability was monitored in real time using the validation set to avoid overfitting. Finally, the diagnostic accuracy of the diagnostic model on the validation set was stabilized at 95%. The optimal model parameters were saved for later use, resulting in a trained diagnostic model. The input data of the diagnostic model consisted of multi-dimensional key feature data (obtained from the test data after preprocessing, covering core characterization indicators such as electrical signals and magnetic signals) after cleaning and normalization. The output data consisted of clearly defined inductive EMC defect categories (such as no defects, excessive electromagnetic leakage, insufficient anti-magnetic interference, etc.) and their corresponding confidence scores. The model parameters were adjusted using the training set, and the model performance (such as accuracy and recall) was evaluated using the validation set until the diagnostic model performance met the standards. The trained diagnostic model was then saved for subsequent diagnosis.

[0036] Once the receiving point collects the test data, it follows the same preprocessing procedure as historical test data, inputting the preprocessed test data into the trained diagnostic model. The trained diagnostic model outputs the corresponding diagnostic results, which can also include a diagnostic confidence level (e.g., "Electromagnetic leakage exceeds the standard, confidence level 96%)". The confidence level can provide a reference for relevant testing personnel. If the confidence level is lower than a preset threshold (e.g., 85%), manual review can be requested to further improve the reliability of the diagnosis.

[0037] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-modal perception based inductive powder formed part EMC defect diagnosis system, characterized in that, include: The model building module is used to define multiple test information for inductor powder molded parts and build corresponding test models based on these test information; the test information includes environmental information and control information. An integration module is used to determine the test plan for inductor powder molded parts. Based on the test plan, multiple test models are integrated to obtain an integrated model. The configuration module is used to configure the cloud server for the integrated model, set multiple collection points and multiple receiving points for the integrated model and the cloud server respectively, and establish a one-to-one transmission channel between the multiple collection points and multiple receiving points. The test execution module is used to test the inductor powder molded parts through the integrated model and transmit the test data collected at the acquisition points to the receiving points through the transmission channel; The diagnostic module is used to preprocess the test data received from multiple receiving points, and input the preprocessed test data into the trained diagnostic model. The trained diagnostic model outputs the diagnostic results of the test data.

2. The multi-modal perception based EMC defect diagnosis system for induction powder formed parts of claim 1, wherein, The process involves defining multiple test parameters for the inductor powder molded parts and constructing corresponding test models based on these parameters, including: Determine the product information of the inductor powder molded part, and formulate the corresponding test information for the inductor powder molded part based on the product information. The test information includes environmental information and control information. A test model is constructed based on the test information corresponding to the inductor powder molded part, and the test information is used as the label information of the corresponding test model.

3. The multi-modal perception based EMC defect diagnosis system for induction powder formed parts of claim 2, wherein, The test plan for determining the inductor powder molded part integrates multiple test models based on the test plan to obtain an integrated model, including: A test plan for inductor powder molded parts is developed based on the product information of the inductor powder molded parts; the test plan includes: the execution order of multiple test models and the sampling frequency of multiple test models; Multiple test models are arranged according to the test plan, and the arranged test models are combined to form an integrated model.

4. The multi-modal perception based EMC defect diagnosis system for induction powder formed parts of claim 1, wherein, The configuration of the cloud server for the integrated model, the setting of multiple collection points and multiple receiving points for both the integrated model and the cloud server, and the establishment of a one-to-one transmission channel between the multiple collection points and multiple receiving points include: Configure a cloud server for multiple test models in the integrated model; Multiple data collection points are deployed for each of the multiple test models in the integrated model. These data collection points are used to collect test data for the corresponding test models. For each collection point, multiple receiving points are deployed within the cloud server. The receiving points are used to receive the test data collected by the collection points through the transmission channel. The number of collection points and receiving points is the same, and a transmission channel is established between the corresponding collection points and receiving points. A distribution center is deployed within the cloud server. The distribution center deploys transmission carriers for corresponding transmission channels based on the collection points. Each transmission carrier is configured with a connection port and given a unique number. Each transmission carrier is spatially divided into a primary data space and multiple secondary data spaces. External storage points and internal storage points are configured for both the primary and secondary data spaces. The internal storage points are located within the external storage points. A first connection point and a second connection point are configured for both the external and internal storage points. The internal storage points are used to store test data collected by the collection points, while the external storage points are used to store spoofed data. A virtual surface is constructed within the cloud server, and multiple acquisition points, receiving points, transmission channels, and transmission carriers are mapped onto the virtual surface to obtain the positions of multiple acquisition points, receiving points, transmission channels, and transmission carriers on the virtual surface; among them, each position is connected one-to-one with the corresponding acquisition point / receiving point / transmission channel / transmission carrier.

5. The multi-modal perception based EMC defect diagnosis system for induction powder formed parts of claim 4, wherein, In the virtual surface: Multiple transmission carriers corresponding to the same acquisition cycle are divided into an adjacent group, and multiple transmission carriers within the adjacent group are topologically connected based on their physical distance. The number of transmission carriers between two transmission carriers in the adjacent group is used as the distance coefficient between the two transmission carriers. The larger the distance coefficient, the farther away the two transmission carriers are, and the smaller the distance coefficient, the closer the two transmission carriers are. Trigger conditions are set for the transmission carrier. The virtual plane monitors the transmission carrier through the trigger conditions and marks the transmission carrier that meets the trigger conditions as an abnormal transmission carrier. The trigger condition is that the transmission carrier is accessed by an unauthorized port.

6. The multi-modal perception based EMC defect diagnosis system for induction powder formed parts of claim 5, wherein, The process of testing inductor powder molded parts using an integrated model and transmitting test data collected from multiple sampling points through a transmission channel includes: Multiple test models within the integrated model are activated to test the inductor powder molded parts. The acquisition points corresponding to each test model collect the test data of the inductor powder molded parts through the acquisition cycle. The data collection point stores the collected test data into the corresponding transmission carrier, which then transmits the data to the corresponding receiving point through the transmission channel. During the transmission process, the virtual surface monitors the transmission carrier in real time.

7. The multi-modal perception based EMC defect diagnosis system for induction powder formed parts of claim 6, wherein, The virtual surface performs real-time monitoring of the transmission medium during transmission, including: When the virtual plane detects that the connection port of the transmission carrier is accessed by an unauthorized port, the transmission carrier is marked as an abnormal transmission carrier. The abnormal transmission carrier will connect the first docking point in the secondary data space to the unauthorized port. Meanwhile, the transmission carrier with the smallest distance coefficient between the adjacent group and the abnormal transmission carrier is selected as the backup transmission carrier. The distribution center sends the same password token to the second docking point of the abnormal transmission carrier and the backup transmission carrier through the point. The abnormal transmission carrier and the backup transmission carrier establish an encrypted transmission channel through the preset connection rules. The encrypted transmission channel transmits the abnormal transmission carrier's number and test data to the internal storage point of the backup transmission carrier's auxiliary data space. The connection rules include establishing an encrypted communication channel between two transmission carriers with the same password token on the virtual plane; When the faulty transmission carrier and the backup transmission carrier arrive at their respective receiving points, the receiving points will send the test data marked with the corresponding number from the backup transmission carrier to the receiving point corresponding to the faulty transmission carrier.

8. The multi-modal perception based EMC defect diagnosis system for induction powder formed parts of claim 1, wherein, The process involves preprocessing test data received from multiple receiving points, inputting the preprocessed test data into a trained diagnostic model, and the trained diagnostic model outputting diagnostic results from the test data, including: Collect and preprocess multiple historical test data and their corresponding actual diagnostic results, construct a historical dataset from the multiple historical test data and their corresponding actual diagnostic results, and use the historical dataset to train the diagnostic model to obtain a trained diagnostic model. The preprocessed test data is input into the trained diagnostic model, which then outputs the diagnostic results corresponding to the test data.

Images