CN115398251A

CN115398251A - Power semiconductor device's test system, cloud server and test machine

Info

Publication number: CN115398251A
Application number: CN202180007976.XA
Authority: CN
Inventors: 杜若阳; 唐诗; 龙纲; 邱辉
Original assignee: Huawei Digital Power Technologies Co Ltd
Current assignee: Huawei Digital Power Technologies Co Ltd
Priority date: 2021-04-30
Filing date: 2021-04-30
Publication date: 2022-11-25
Also published as: WO2022227018A1

Abstract

A test system, a cloud server and a tester of a power semiconductor device are provided, the system comprises: a cloud server (100) and at least one tester (200); the cloud server (100) is used for training by utilizing historical data of a plurality of power semiconductor devices in advance to obtain an artificial intelligence model; a tester (200) for obtaining data of a power semiconductor device to be tested; the cloud server (100) or the tester (200) is used for inputting data of the power semiconductor device to be tested into the artificial intelligence model, and the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested. The cloud server (100) or the testing machine (200) diagnoses the power semiconductor device to be tested by using the trained artificial intelligence model, so that an accurate diagnosis result can be obtained, and abnormal power semiconductor devices are removed, so that the whole circuit is prevented from being failed due to the fact that the failure occurs in the using process of an actual product.

Description

Power semiconductor device's test system, cloud server and test machine

Technical Field

The application relates to the technical field of power semiconductors, in particular to a test system, a cloud server and a test machine for a power semiconductor device.

Background

The power semiconductor device generally refers to a semiconductor device that realizes a circuit switching function, and may be applied to a power conversion circuit of an electric vehicle, a photovoltaic system, or a data center, for example. For example, in an electric vehicle, the power semiconductor device may be applied to a powertrain control circuit of the electric vehicle. In practical applications, in order to ensure the safety of the circuit to which the power semiconductor device is applied, the power semiconductor device needs to be subjected to performance testing after production to reduce the risk of failure of the power semiconductor device.

At present, the quality test of the power semiconductor device mainly sets upper and lower limit values for performance parameters of the power semiconductor device through a testing machine, and the power semiconductor device with the performance parameters exceeding the upper and lower limit values is eliminated.

However, the existing method for testing by using the upper and lower limit values needs to be confirmed by a large number of historical products, and the screening method by using the upper and lower limit values has high dependence on experience, cannot establish a complete association relation with the failure principle of the device, and can only identify a part of power semiconductor devices with problems.

Disclosure of Invention

The application provides a test system, a cloud server and a test machine for power semiconductor devices, which can more comprehensively screen bad power semiconductor devices.

The test system of the power semiconductor device provided by the embodiment of the application comprises a cloud server and at least one test machine; the cloud server is used for training by utilizing historical data of a plurality of power semiconductor devices in advance to obtain an Artificial Intelligence (AI) model; the tester is used for obtaining data of the power semiconductor device to be tested; the cloud server or the testing machine can diagnose the power semiconductor device to be tested by utilizing the trained artificial intelligence model, namely, the cloud server or the testing machine inputs data of the power semiconductor device to be tested into the artificial intelligence model, and the output of the artificial intelligence model is a diagnosis result of the power semiconductor device to be tested.

According to the scheme provided by the embodiment of the application, the powerful operation energy and the storage capacity of the cloud server are utilized to train to obtain the accurate artificial intelligence model, then the high-risk power semiconductor device in the power semiconductor device to be tested can be accurately diagnosed by utilizing the pre-trained artificial intelligence model, so that the limitation that the traditional test system only uses upper limit data to screen the high-risk power semiconductor device is made up, and whether the risk exists in the power semiconductor device can be diagnosed more accurately and comprehensively.

In a possible implementation manner, in the embodiment of the application, the cloud server is configured to input data of the power semiconductor device to be tested into the artificial intelligence model, output of the artificial intelligence model is a diagnosis result of the power semiconductor device to be tested, and send the diagnosis result to the test machine. It should be understood that, in the embodiment of the present application, the cloud server inputs data into the artificial intelligence model to obtain a diagnosis result, and sends the diagnosis result to the testing machine, so that the testing machine can directly obtain a test result, and the computing resource of the testing machine is saved.

In a possible implementation manner, the cloud server in the embodiment of the application is used for sending the pre-trained artificial intelligence model to the testing machine; the testing machine inputs the data of the power semiconductor device to be tested into the artificial intelligence model, and the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested. It should be appreciated that the cloud server sends the trained artificial intelligence model to the tester, which then inputs the data and obtains the diagnostic result. Therefore, the test machine completes the test work of the power semiconductor device to be tested by a single machine, the diagnosis result does not need to be obtained from the cloud server, the dependence on the network between the cloud server and the test machine is reduced, and the instantaneity of the test system of the power semiconductor device to be tested in the embodiment of the application in obtaining the diagnosis result is improved.

In a possible implementation manner, the testing machine in the embodiment of the present application is further configured to collect historical data of the plurality of power semiconductor devices, and send the collected historical data of the plurality of power semiconductor devices to the cloud server; the historical data comprises at least one item of chip test data of a plurality of power semiconductor devices, packaged module function test data or related data when actual application is abnormal; the cloud server is further used for training by using historical data of the plurality of power semiconductor devices by using at least one of a supervised learning model or an unsupervised learning model to obtain an artificial intelligence model.

In a possible implementation mode, a cloud server utilizes an unsupervised learning model to train and obtain an artificial intelligence model, and is specifically used for obtaining failure types of a plurality of power semiconductor devices, all test items corresponding to each failure type are extracted from test items of the power semiconductor devices by utilizing expert knowledge, all test items corresponding to each failure type form a test item subset, each test item subset is distinguished from an abnormal subset by utilizing the unsupervised learning model, a first score and a second score are respectively recorded on the abnormal subset and the normal subset, and the sum of the first score and the second score of all test item subsets of each power semiconductor device is obtained as an abnormal level total score; and when the total abnormal level score is greater than or equal to a preset score threshold value, judging that the power semiconductor device is an abnormal device. It should be understood that the embodiment of the present application, which trains the AI model by using the unsupervised learning model, can make full use of expert knowledge, perform anomaly detection on a subset of test items including a plurality of test items together, the risk of failure of the power semiconductor device can be detected from multiple dimensions by using the relation between devices in the same test item subset or the physical basis corresponding to the same failure type, so that the risk of failure of the device can be more accurately evaluated.

In one possible implementation form of the method, the cloud server in the embodiment of the application is specifically configured to use expert knowledge to mark different first scores for abnormal devices in different test item subsets. It should be understood that different subsets of test items may have different weights for determining abnormal devices, and a more accurate diagnosis result may be obtained by assigning different first scores to the abnormal devices in different subsets of test items.

In a possible implementation manner, in the embodiment of the application, the cloud server obtains the artificial intelligence model by using supervised learning model training, and is specifically configured to obtain relevant data when the power semiconductor device is abnormal as a data tag, extract data features of the data tag by using the supervised learning model, obtain an abnormal level total score of each power semiconductor device by using the data features, and determine that the power semiconductor device is an abnormal device when the abnormal level total score is greater than a preset score threshold.

In a possible implementation manner, the testing machine in the embodiment of the present application is further configured to compare data of the power semiconductor device to be tested with preset upper and lower limit values, and determine that the power semiconductor device to be tested is an abnormal device when the data of the power semiconductor device to be tested exceeds the upper and lower limit values. It should be understood that the test system for the power semiconductor device provided in the embodiment of the present application may combine a conventional test method with an artificial intelligence model, and perform secondary detection by using the artificial intelligence model in the embodiment of the present application on the basis of the conventional test method, so as to improve the accuracy of detection, achieve the complementation between the conventional mode and the AI model test mode, and more comprehensively diagnose the high-risk power semiconductor device.

The embodiment of the application does not limit the number of the testers included in the test system, and the test system may include one tester or two or more testers. When the test system comprises a plurality of test machines, the cloud server may train a global AI model, and then obtain a local AI model suitable for each test machine based on the global AI model according to differences between the test machines. That is, in a possible implementation manner, the test system provided in the embodiment of the present application at least includes the following two test machines: a first tester and a second tester; the cloud server is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, adjusting the global artificial intelligence model by using the first historical data to obtain a first artificial intelligence model, and adjusting the global artificial intelligence model by using the second historical data to obtain a second artificial intelligence model; the method comprises the steps of testing a power semiconductor device to be tested corresponding to a first testing machine by using a first artificial intelligence model to obtain a first diagnosis result, testing the power semiconductor device to be tested corresponding to a second testing machine by using a second artificial intelligence model to obtain a second diagnosis result, sending the first diagnosis result to the first testing machine, and sending the second diagnosis result to the second testing machine. According to the embodiment of the application, the global artificial intelligence model is trained by utilizing the first historical data and the second historical data, all historical data can be fully utilized, and the test accuracy of the global artificial intelligence model is improved. On the basis, considering that the test performances of different test machines may have differences, the application also adjusts the global artificial intelligence model by using the first historical data to obtain the first artificial intelligence model, and adjusts the global artificial intelligence model by using the second historical data to obtain the second artificial intelligence model, so that the local artificial intelligence model (the first artificial intelligence model) aiming at the characteristics of the first test machine and the local artificial intelligence model (the second artificial intelligence model) aiming at the characteristics of the second test machine can be obtained, and the test accuracy of the artificial intelligence model used in the actual test process is further improved.

In this embodiment, each tester corresponds to a global AI model, and the cloud server obtains a diagnosis result. That is, in a possible implementation manner, the test system provided in the embodiment of the present application at least includes the following two test machines: a first tester and a second tester; the cloud server is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, testing the power semiconductor device to be tested corresponding to the first testing machine by using the global artificial intelligence model to obtain a first diagnosis result, testing the power semiconductor device to be tested corresponding to the second testing machine by using the global artificial intelligence model to obtain a second diagnosis result, sending the first diagnosis result to the first testing machine, and sending the second diagnosis result to the second testing machine. The cloud server provided by the embodiment of the application utilizes the first historical data and the second historical data to train the global artificial intelligence model, and can make full use of all historical data, so that the generalization capability of the global artificial intelligence model is improved, the global artificial intelligence model with higher accuracy is utilized to diagnose the power semiconductor device to be tested, the universality of the artificial intelligence model can be improved, and the accuracy of new device diagnosis is increased.

In this embodiment, the cloud server obtains the local AI models corresponding to the respective testing machines and sends the local AI models to the respective testing machines, and the respective testing machines complete diagnosis. That is, in a possible implementation manner, the test system for a power semiconductor device in the embodiment of the present application includes at least the following two test machines: a first tester and a second tester; the cloud server is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, adjusting the global artificial intelligence model by using the first historical data to obtain a first artificial intelligence model and sending the first artificial intelligence model to the first testing machine, and adjusting the global artificial intelligence model by using the second historical data to obtain a second artificial intelligence model and sending the second artificial intelligence model to the second testing machine; the first testing machine is specifically used for diagnosing the corresponding power semiconductor device to be tested by utilizing the first artificial intelligent model; and the second testing machine is specifically used for diagnosing the corresponding power semiconductor device to be tested by using the second artificial intelligence model. The testing machine of the embodiment of the application can directly utilize the artificial intelligence model, a single machine completes the task of detecting the power semiconductor device to be tested, and communication with the cloud server at the cloud end is not needed, so that when the network between the testing machine at the board end and the cloud server at the cloud end is interrupted, the testing system of the power semiconductor device in the embodiment of the application can not obtain the diagnosis result of the power semiconductor device to be tested. Therefore, the test system for the power semiconductor device provided by the embodiment of the application can reduce the network dependence between the cloud server and the tester, and improve the real-time performance of the test system for the power semiconductor device for obtaining the diagnosis result in the embodiment of the application.

In this embodiment, the cloud server sends the global AI model to each tester, and each tester completes diagnosis by using the global AI model. That is, in a possible implementation manner, the test system for a power semiconductor device in the embodiment of the present application includes at least the following two test machines: a first tester and a second tester; the cloud server is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, and sending the global artificial intelligence model to the first testing machine and the second testing machine; the first testing machine is used for diagnosing the corresponding power semiconductor device to be tested by utilizing the global artificial intelligence model; and the second testing machine is used for diagnosing the corresponding power semiconductor device to be tested by utilizing the global artificial intelligence model. The testing machine in the embodiment of the application can finish the testing work of the power semiconductor device to be tested by a single machine according to the trained global artificial intelligence model. When the test machine obtains the diagnosis result, the cloud server does not need to issue the diagnosis result to the test machine, so that the test machine obtains the diagnosis result without being influenced by network faults and has low dependence on the network.

In a possible implementation manner, the testing machine in the embodiment of the present application is further configured to send update data to the cloud server; the cloud server is also used for updating the artificial intelligence model according to the updating data. In addition, the tester is also used for finely adjusting the pre-trained AI model by combining the data of the tester side, diagnosing the power semiconductor device to be tested by using the model after fine adjustment and outputting the diagnosis result. It should be understood that, in the embodiment of the present application, the tester may perform fine adjustment on the obtained AI model based on its own data or characteristics of the data, so as to further improve the accuracy of the AI model in the embodiment of the present application. The tester can send the fine-tuned AI model to a cloud server, and the cloud server performs unified management.

Based on the test system for the power semiconductor device provided by the embodiment, the embodiment of the application also provides a cloud server, and the advantages of the test system in each embodiment are also applicable to the following servers, and are not described again. The server includes: a first transceiver device and a first controller; the first transceiver is used for receiving test data of the power semiconductor device to be tested, which are sent by the test machine; the first controller is used for training by utilizing historical data of a plurality of power semiconductor devices in advance to obtain an artificial intelligence model; the system is also used for inputting the data of the power semiconductor device to be tested into an artificial intelligence model, and the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested; the first transceiver is also used for sending the diagnosis result to the tester; or the first controller is used for sending the artificial intelligence model to the testing machine so that the testing machine diagnoses the power semiconductor device to be tested by using the artificial intelligence model.

In a possible manner, the first controller in the embodiment of the present application is specifically configured to, when obtaining the artificial intelligence model by using unsupervised learning model training, specifically to obtain failure types of a plurality of power semiconductor devices, extract all test items corresponding to each failure type in the test items of the power semiconductor devices by using expert knowledge, all test items corresponding to each failure type form a test item subset, abnormal level detection is carried out on each test item subset by using an unsupervised learning model, a first score and a second score are respectively recorded on an abnormal device and a normal device, and the sum of the first score and the second score of all test item subsets of each power semiconductor device is obtained and used as an abnormal level total score; and when the total abnormal level score is greater than a preset score threshold value, judging that the power semiconductor device is an abnormal device.

In a possible manner, the cloud server in the embodiment of the present application corresponds to at least two of the following test machines: a first tester and a second tester; the first controller is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, adjusting the global artificial intelligence model by using the first historical data to obtain a first artificial intelligence model, and adjusting the global artificial intelligence model by using the second historical data to obtain a second artificial intelligence model; testing the power semiconductor device to be tested corresponding to the first testing machine by using the first artificial intelligent model to obtain a first diagnosis result, and testing the power semiconductor device to be tested corresponding to the second testing machine by using the second artificial intelligent model to obtain a second diagnosis result; the first transceiver is specifically configured to send the first diagnostic result to the first tester, and is further configured to send the second diagnostic result to the second tester.

In a possible manner, the cloud server in the embodiment of the present application corresponds to at least two of the following test machines: a first tester and a second tester; the first controller is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, testing the power semiconductor device to be tested corresponding to the first testing machine by using the global artificial intelligence model to obtain a first diagnosis result, and testing the power semiconductor device to be tested corresponding to the second testing machine by using the global artificial intelligence model to obtain a second diagnosis result; the first transceiver is specifically configured to send the first diagnostic result to the first tester, and is further configured to send the second diagnostic result to the second tester.

In a possible manner, the first transceiver device in the embodiment of the present application is further configured to receive update data sent by the tester; the first controller is also used for updating the artificial intelligence model according to the updating data.

Based on the test system and the server for the power semiconductor device provided by the above embodiments, the embodiments of the present application also provide a testing machine, and the advantages of the above test system in each embodiment are also applicable to the following testing machine, which are not described again. This test machine includes: a second transceiver device and a second controller; the second transceiver is used for receiving the artificial intelligence model sent by the cloud server, and the artificial intelligence model is obtained by training the cloud server by using historical data of the plurality of power semiconductor devices in advance; the second controller is used for obtaining data of the power semiconductor device to be tested; and inputting the data of the power semiconductor device to be tested into the artificial intelligence model, wherein the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested.

In a possible manner, the second controller in the embodiment of the present application is further configured to collect historical data of the plurality of power semiconductor devices, and send the collected historical data of the plurality of power semiconductor devices to the cloud server; the historical data includes at least one of chip test data of the plurality of power semiconductor devices or module function test data after packaging.

In a possible manner, the second controller in this embodiment is further configured to compare the data of the power semiconductor device to be tested with preset upper and lower limit values, and determine that the power semiconductor device to be tested is an abnormal device when the data of the power semiconductor device to be tested exceeds the upper and lower limit values.

According to the technical scheme, the embodiment of the application has the following advantages:

the test system provided by the embodiment of the application comprises the cloud server and the test machine, the computing capacity and the storage capacity of the cloud server are higher than those of the test machine, and therefore the artificial intelligence model is obtained by utilizing the powerful computing capacity and the powerful storage capacity of the cloud server through pre-training. The cloud server trains the artificial intelligence model by utilizing a large amount of historical data of the power semiconductor devices in advance, so that parameters of the artificial intelligence model can be accurately obtained, the cloud server or the testing machine can accurately diagnose the power semiconductor devices to be tested by utilizing the trained artificial intelligence model, accurate diagnosis results are obtained, abnormal power semiconductor devices are removed, and the problem that faults occur in the using process of actual products to cause the faults of the whole circuit is avoided. The testing system can make up for the defects of a traditional mode, namely the traditional mode only utilizes upper and lower limit values to screen the high-risk power semiconductor devices, and some high-risk power semiconductor devices cannot be screened out. The scheme can make up the inaccuracy of a testing machine for simply screening defective products of the power semiconductor device by utilizing the upper limit value and the lower limit value.

Drawings

Fig. 1 is an architecture diagram of a test system for a power semiconductor device according to an embodiment of the present disclosure;

fig. 2 is a schematic diagram illustrating a data source of a power semiconductor device according to an embodiment of the present disclosure;

fig. 3A is a flowchart of a method for testing a power semiconductor device according to an embodiment of the present disclosure;

fig. 3B is a flowchart of another power semiconductor device testing method according to an embodiment of the present disclosure;

fig. 4 is a schematic diagram of another power semiconductor device testing system provided in an embodiment of the present application;

fig. 5 is a schematic diagram of a test system for a power semiconductor device according to an embodiment of the present disclosure;

fig. 6 is a schematic diagram of another power semiconductor device testing system provided in an embodiment of the present application;

FIG. 7 is a schematic diagram of a testing system for a power semiconductor device according to an embodiment of the present application;

fig. 8 is a schematic diagram of a test system for a power semiconductor device including multiple testers according to an embodiment of the present disclosure;

fig. 9 is a schematic diagram of another testing system for a power semiconductor device including multiple testers according to an embodiment of the present disclosure;

fig. 10 is a schematic diagram of a testing system for a power semiconductor device including a plurality of testers according to another embodiment of the present disclosure;

fig. 11 is a schematic diagram of a testing system for a power semiconductor device including a plurality of testers according to another embodiment of the present disclosure;

fig. 12 is a schematic structural diagram of a cloud server according to an embodiment of the present application;

fig. 13 is a schematic structural diagram of a testing machine according to an embodiment of the present application.

Detailed Description

The terms "first," "second," and the like in the following description are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of technical features indicated. Thus, a feature defined as "first," "second," etc. may explicitly or implicitly include one or more of that feature. In the description of the present application, unless otherwise indicated, "plurality" means two or more.

Further, in the present application, the terms "upper", "lower", and the like may include, but are not limited to, being defined relative to the orientation in which the components in the drawings are schematically disposed, and it is to be understood that these directional terms may be relative concepts that are used for descriptive and clarifying purposes relative to the orientation in which the components in the drawings are disposed, and that will vary accordingly.

In the present application, unless expressly stated or limited otherwise, the term "coupled" is to be construed broadly, e.g., "coupled" may be a fixed connection, a removable connection, or an integral part; may be directly connected or indirectly connected through an intermediate. Furthermore, the term "coupled" may be a manner of making electrical connections that communicate signals. "coupled" may be a direct electrical connection or an indirect electrical connection through intervening media.

The embodiment of the application relates to a test system of a power semiconductor device, wherein the test system of the power semiconductor device comprises: cloud server and at least one test machine. The cloud server has strong computing energy and storage space, so the cloud server can train the artificial intelligence model by utilizing a large amount of historical data of the power semiconductor devices. Because the AI module is obtained by training a large amount of historical data, the cloud server or the testing machine can accurately diagnose the power semiconductor device to be tested by using the trained AI model, and the abnormal power semiconductor device is removed, so that the whole circuit is prevented from being failed in the real use process. The scheme can solve the problem that the inaccuracy of screening the defective products of the power semiconductor device to be tested is simply realized by only utilizing the upper limit value and the lower limit value preset by the testing machine in the traditional scheme.

In order to enable those skilled in the art to better understand the technical solutions provided by the embodiments of the present application, a test system provided by the embodiments of the present application is described in detail below with reference to the accompanying drawings, where the test system includes a cloud server and at least one tester.

The number of the testing machines is not particularly limited in the embodiment of the application, and the testing machines can be set according to the number of actual product lines, one testing machine can correspond to one product line, and one product line can also correspond to a plurality of testing machines. The types of the power semiconductor devices corresponding to the plurality of product lines may be the same or different. While three test machines are described below with reference to fig. 1, it should be understood that one test machine may be used, two test machines may be used, or more test machines may be used.

System embodiment

Referring to fig. 1, a diagram of a test system architecture of a power semiconductor device according to an embodiment of the present application is shown.

As shown in fig. 1, the test system for a power semiconductor device according to the embodiment of the present disclosure includes a cloud server 100, a first test machine 201, a second test machine 202, and a third test machine 203. The cloud service 100 is located on a cloud platform, and the cloud platform may include a plurality of cloud servers, and only one cloud server is illustrated in the figure as an example. The first testing machine 201, the second testing machine 202 and the third testing machine 203 at the board end can respectively communicate with the cloud server 100 at the cloud end for data exchange or transmission.

The cloud server 100 is configured to perform training in advance by using historical data of a plurality of power semiconductor devices to obtain an AI model.

It should be understood that the embodiment of the present application is not particularly limited to the specific model used by the cloud server 100 for AI model training, and for example, a supervised learning model may be used, and an unsupervised learning model may also be used. The historical data used by the cloud server 100 to train the AI model may be data uploaded by the first testing machine 201, the second testing machine 202, and the third testing machine 203, or may also be data uploaded by more testing machines, which is not specifically limited in this embodiment. It should be understood that, when the cloud server 100 performs AI model training, the more history data is used, the more accurate the parameters of the trained AI model are.

The first tester 201, the second tester 202 and the third tester 203 are used for obtaining data of the power semiconductor device to be tested corresponding to each.

During specific testing, the cloud server 100 may perform the testing, or the cloud server 100 issues the trained AI model to the testing machine, and the testing machine performs the testing.

First, the cloud server 100 performs a test.

The cloud server 100 is configured to input data of the power semiconductor device to be tested into a pre-trained artificial intelligence model, and output of the artificial intelligence model is a diagnosis result of the power semiconductor device to be tested. And sending the diagnosis result to the testing machine.

It should be understood that the power semiconductor device to be tested herein refers broadly to any power semiconductor device that needs to be tested; a tester herein broadly refers to any tester.

And the second method comprises the following steps: the testing machine performs the test.

The first tester 201, the second tester 202 and the third tester 203 are respectively used for inputting the data of the power semiconductor device to be tested corresponding to each tester into the artificial intelligence model trained in advance, and respectively obtaining the output of the model trained in advance as the diagnosis result of the power semiconductor device to be tested corresponding to each tester.

The embodiment of the application does not limit whether the first tester 201, the second tester 202 and the third tester 203 test the corresponding power semiconductor devices to be tested at the same time or respectively and sequentially, because the testing actions of each tester can be completed independently and do not interfere with each other.

It can be understood that, in order to improve the accuracy of intercepting defective products of the power semiconductor device, the performance of the power semiconductor device is tested by using the artificial intelligence model. However, because the computation capability and the storage capability of the board-end tester are limited, if the board-end tester is used to train the artificial intelligence model, the time consumption is long, and excessive computation resources are occupied, thereby affecting other functions of the tester. Therefore, the power semiconductor device testing system provided by the embodiment of the application utilizes the strong computing capability and storage capability of the cloud server, obtains the artificial intelligence model through the historical data training of the plurality of power semiconductor devices, and then detects the power semiconductor device to be tested through the artificial intelligence model. The AI model trained by the cloud server can be sent to a testing machine, the testing machine tests the power semiconductor device to be tested by using the AI model trained by the cloud server, and finally, a diagnosis result is output.

Therefore, the test system of the power semiconductor device provided by the embodiment of the application can diagnose the performance of the power semiconductor by using the artificial intelligence model and screen the defective products in the power semiconductor device. In addition, considering that the testing machine at the board end is limited in computing capacity and storage capacity, the cloud server with strong computing capacity and storage capacity is used for training the artificial intelligence model, and therefore the artificial intelligence model with high diagnosis accuracy is obtained.

The foregoing mainly introduces the architecture of the power semiconductor device test system provided in the embodiment of the present application, and the following introduces a specific process of performing AI model training by the cloud server with reference to the accompanying drawings.

Referring to fig. 2, a data source diagram of a power semiconductor device according to an embodiment of the present disclosure is shown.

In the embodiment of the application, the cloud server utilizes a large amount of historical data for training to obtain the AI model. The historical data may include production line data and application side data.

The source of the production line data D1 may include data obtained by the following tests: chip (CP) testing, functional Testing (FT), single board testing, and overall testing. The CP test may be a test of a single chip, and the FT test may be a test of a packaged module, which has a certain circuit function. The single board test may be a board-on-board test, and the whole machine test may be an on-machine test. It can be understood that the chip test data and the power test data can be used separately for AI model training, or can be used together for AI model training.

It should be noted that the CP test is mainly performed on a single chip, and the single chip may be a chip of an isolated Insulated Gate Bipolar Transistor (IGBT), an isolated Metal-Oxide Semiconductor Field Effect Transistor (MOSFET), a power Semiconductor device such as a diode (diode), a triode (Bipolar Junction Transistor), a Thyristor (Thyristor), an Integrated Gate-Commutated Thyristor (IGCT), or a chip formed by a plurality of IGBTs, or may be a chip formed by an IGBT and a diode. The FT test is mainly for one packaged module,

the application side data D2 mainly includes relevant data corresponding to the power semiconductor device after being produced and when failing in the actual use process, that is, the actual working condition data of the power semiconductor device fed back by the client side may include a test data source or a verification data source

As an example, the test data of the power semiconductor device to be tested in the embodiment of the present application may include, but is not limited to, the following breakdown voltage, leakage current, turn-on voltage drop, parasitic capacitance, gate level charge, turn-on loss, and turn-off loss.

The following describes the principle of the power semiconductor device test system provided by the present application for obtaining a diagnostic result using test data.

The AI model obtained by cloud server training can be obtained by using unsupervised integrated learning model training or by using supervised learning model training; or obtained by using a combination of a supervised learning model and an unsupervised learning model, which is not specifically limited in the embodiments of the present application

The following first introduces the embodiments of the present application to provide that the AI model is obtained by unsupervised ensemble learning model training.

As an example, the unsupervised learning model in the embodiment of the present application may be an unsupervised ensemble learning model. It can be understood that the unsupervised model in the embodiment of the application adopts an unsupervised learning algorithm and expert knowledge to construct an unsupervised ensemble learning model, so that expert experience can be fully utilized, and the influence of too few failed samples on the accuracy of the artificial intelligent model is reduced.

In the embodiment of the application, when the cloud server obtains the artificial intelligence model by using the unsupervised learning model for training, as a possible implementation mode, the cloud server is specifically used for obtaining failure types of a plurality of power semiconductor devices, all test items corresponding to each failure type are extracted from the test items of the power semiconductor devices by using expert knowledge, all test items corresponding to each failure type form a test item subset, each test item subset is distinguished from an abnormal subset by using the unsupervised learning model, a first score and a second score are respectively recorded for the abnormal subset and the normal subset, and the sum of the first score and the second score of all test item subsets of each power semiconductor device is obtained as an abnormal level total score; and when the total abnormal level score is greater than or equal to a preset score threshold value, judging that the power semiconductor device is an abnormal device.

The failure type may be various failure types that cannot be intercepted by the card control during the single test of the power semiconductor device. For example, power semiconductor devices have early failures in actual use, but are not screened out during controlled testing.

In addition, a failure type may be problematic for many test items, such as gate leakage, may be a problem at the physical layer, and many test items involved may be caused by a problem at the physical layer. Therefore, all the test items associated with one failure type are extracted to form the test item subset corresponding to the test item.

As an example, the following specifically describes, with reference to table 1 and table 2, a scheme for obtaining an artificial intelligence model by using unsupervised ensemble learning model training, which is provided in an embodiment of the present application.

TABLE 1

TABLE 2

	Subset1	Subset2	Subset3	……	Total score of abnormal level
						Device A	0	0	0	0
Device B	0	1	0		1
						Device C	0	0	0	0
……					……

The Gate failure, the PN junction failure or the pressure ring failure are different failure types respectively; IGES (GE leakage current), VTH (threshold voltage), VCESAT (conduction voltage drop between the C pole and the E pole), ICES (CE leakage current) and Eon (turn-on loss) etc. are different test items, as can be seen from Table 1, different failure types include different test items, and every failure type includes a plurality of test items, therefore, this application embodiment is when the AI model training, the data that use are multidimensional data, and is not traditional one-dimensional data, therefore, the AI model that the training obtained is more accurate, and is more comprehensive, when finally utilizing the AI model of training to diagnose, the diagnostic result of output is also more accurate.

Wherein, the device A, the device B and the device C are different power semiconductor devices respectively. Subset1, subset2 and Subset3 are respectively different test item subsets; i.e., each device includes at least three subsets of test items. Each subset of test items includes a different content and number of test items. In the table, the second score of the abnormal subset is 1, the first score of the normal subset is 0, and the sum of the first score and the second score of all the test item subsets of each power semiconductor device is obtained as the total score of the abnormal level; that is, the abnormal level of the device a is always 0, the abnormal level of the device B is always 1, and the abnormal level of the device C is always 0. And if the total abnormal level score is greater than or equal to a preset score threshold value, judging that the power semiconductor device is an abnormal device. For example, if the preset score threshold is 1, the test device B is an abnormal device.

Referring to fig. 3A, the figure is a flowchart of a testing method of a power semiconductor device according to an embodiment of the present application.

The test method for the power semiconductor device provided by the embodiment of the application comprises the following steps:

s301: failure types of the plurality of power semiconductor devices are obtained.

As an example, failure types such as Gate failure, PN junction failure, or pressure ring failure may be obtained. It can be understood that the more types of failures are collected, the more comprehensive the more historical data of the same power semiconductor device.

S302: and extracting all test items corresponding to each failure type from the test items of the power semiconductor device by using expert knowledge, wherein all test items corresponding to each failure type form a test item subset.

In this example, the Gate failures correspond to IGES, VTH, VCESAT, and Eon, among other test items, which form a subset of test items subset1. The PN junction failures correspond to test entries such as VCESAT and ICES, which form a subset of test entries subset2. The failure of the pressure ring corresponds to the test items such as ICES, and the test items form a test item subset3.

S303: and distinguishing the abnormal subset and the normal subset of each test item subset by using an unsupervised learning model to obtain a first score and a second score of the abnormal subset and the normal subset respectively.

For example, abnormal point detection is performed on the test item subset corresponding to the Gate failure by using an unsupervised learning model, and the obtained normal device A, the normal device B and the normal device C are respectively recorded with second scores. And carrying out abnormal point detection on the test item subset corresponding to the PN junction failure by using an unsupervised learning model, and respectively recording a second score on the obtained normal device A and a first score on the obtained normal device C, and recording a first score on the obtained abnormal device B. In the example of table 2, the first score is 1 and the second score is 0.

In practical applications, in order to more accurately evaluate the risk of failure of each device, as a possible implementation manner, the cloud server further uses expert knowledge to record different first scores for abnormal devices in different test item subsets. I.e. weights corresponding to the subset of test items are added on the basis of the first score and the second score. For example, there are three failure types in total, the three failure types correspond to three subsets of test items a, b, and c, the test subset a is 0.2, the test subset b is 0.3, and the test subset c is 0.5. Accordingly, if device a and device B in the B-test subset are normal, then device a and device B, respectively, may be scored a second score of 0. Device C is abnormal and may note the initial score of 1*b with a weight of 0.3= the first score of 0.3 for the test subset.

S304: and obtaining the sum of the first score and the second score of all the test item subsets of each power semiconductor device as the total abnormal level score.

In the corresponding examples of tables 1 and 2, the abnormality level of device a is totally divided into 0 as the sum of the second score 0 of test item subset1, the second score 0 of test item subset2, and the second score 0 of test item subset3. The abnormality level of the device B is always divided into the sum of the second score 0 of the test item subset1, the first score 1 of the test item subset2, and the second score 0 of the test item subset3, and is 1. The abnormality level of the device C is always divided into the sum of the second score 0 of the test item subset1, the second score 0 of the test item subset2, and the second score 0 of the test item subset3, and is 0.

S305: and when the total abnormal level score is greater than or equal to a preset score threshold value, judging that the power semiconductor device is an abnormal device.

In this example, the preset score threshold may be 1, and of course, other values may be obtained according to actual situations, and the application is not limited herein. When the preset score threshold is 1, the device B is judged to be an abnormal device because the total abnormal level of the device B is 1 or more than or equal to the preset score threshold.

Therefore, each failure type corresponds to a plurality of test items, the test item subsets comprising the test items are subjected to abnormal point detection together, and the risks of failure of the power semiconductor device can be detected from multiple dimensions by using the relation between devices in the same test item subset or the physical basis corresponding to the same failure type, so that the risks of failure of the device can be evaluated more accurately.

The scheme described in the above embodiment is that the cloud server obtains the artificial intelligence model through unsupervised learning model training, and as another possible implementation, the cloud server obtains the artificial intelligence model through supervised learning model training is described next.

Referring to fig. 3B, the figure is a flowchart of another power semiconductor device testing method provided in the embodiments of the present application.

The method for testing the power semiconductor device provided by the embodiment of the application comprises the following steps:

s311: and obtaining relevant data when the power semiconductor device is abnormal as a data tag. For example, a very small number of failure samples may be selected as data tags.

S312: and extracting the data characteristics of the data labels by using a supervised learning model.

S313: and obtaining the abnormal level total score of each power semiconductor device by using the data characteristics.

S314: and when the total abnormal level score is larger than a preset score threshold value, judging that the power semiconductor device is an abnormal device. Therefore, the scheme provided by the embodiment of the application can obtain the artificial intelligence model for diagnosing the power semiconductor device through the supervised learning model or the unsupervised learning model.

It should be understood that the method for screening the defective products of the power semiconductor device by using the artificial intelligence model in the embodiment of the present application is a scheme improved on the basis of the conventionally set upper and lower limit numerical diagnosis method. The testing machine is used for diagnosing the power semiconductor device to be tested firstly by utilizing preset upper and lower limit values, namely the testing machine is also used for comparing the data of the power semiconductor device to be tested with the preset upper and lower limit values, and when the data of the power semiconductor device to be tested exceeds the upper and lower limit values, the power semiconductor device to be tested is judged to be an abnormal device. And then the test machine or the cloud server diagnoses the power semiconductor device to be tested by utilizing the AI model, and the diagnosis results of the two schemes are subjected to logic OR operation. Therefore, the test system provided by the embodiment of the application can complement the two schemes, so that the power semiconductor device to be tested can be more comprehensively diagnosed.

It can be understood that, the test system for the power semiconductor device provided in the embodiment of the present application may compare the data of the power semiconductor device to be tested with the preset upper and lower limit values by using a conventional method while diagnosing the power semiconductor device by using the artificial intelligence model, so as to diagnose the power semiconductor device. As a possible implementation manner, when at least one of the scheme of the artificial intelligence model and the scheme of the upper and lower limit values diagnoses that the power semiconductor device fails, the power semiconductor device is judged to be failed, that is, the two diagnosis results are logically or-operated. Therefore, the test system of the power semiconductor device provided by the embodiment of the application can combine the traditional test method with the artificial intelligence model, and further improve the accuracy of detection.

For example, the data of the power semiconductor device X does not exceed the preset upper and lower limit data, but after the data of the power semiconductor device X is input into the artificial intelligence model, the diagnosis result output by the artificial intelligence model is failure, and at this time, it is determined that the power semiconductor device X is failed. And after the data of the power semiconductor device Y exceeds the preset upper and lower limit data, and the data of the power semiconductor device Y is input into the artificial intelligence model, the diagnosis result output by the artificial intelligence model is non-failure, and at the moment, the power semiconductor device Y is judged to be failed. And after the data of the power semiconductor device Z exceeds the preset upper and lower limit data and is input into the artificial intelligence model, the diagnosis result output by the artificial intelligence model is failure, and at the moment, the power semiconductor device Z is judged to be failed.

In the test system for the power semiconductor device provided by the embodiment of the application, in order to make the diagnosis of the artificial intelligence model more accurate, as a possible implementation manner, the test machine is further used for sending update data to the cloud server; and the cloud server is also used for updating the artificial intelligence model according to the updating data.

It should be noted that, in the embodiment of the present application, the step of inputting the data of the power semiconductor device to be tested into the pre-trained artificial intelligence model and obtaining the diagnosis result may be completed by the cloud server at the cloud end or by the testing machine at the board end, and the embodiment of the present application is not limited herein. Two different implementations will be specifically described below by way of examples.

Referring to fig. 4, the figure is a schematic diagram of another power semiconductor device testing system provided in the embodiment of the present application.

The test machine 200 is further configured to collect historical data of the plurality of power semiconductor devices and send the collected historical data of the plurality of power semiconductor devices to the cloud server 100, and the cloud server 100 may train the artificial intelligence model through the historical data of the plurality of power semiconductor devices.

The cloud server 100 is further configured to obtain an artificial intelligence model by training using at least one of a supervised learning model or an unsupervised learning model using historical data of the plurality of power semiconductor devices.

In the embodiment of the present application, the historical data may include at least one of chip test data of the plurality of power semiconductor devices, module function test data after packaging, or related data when the actual application is abnormal. Specifically, as one possible implementation, the historical data may include one or more of the training data in fig. 2.

It should be noted that the artificial intelligence model obtained by training in the embodiment of the present application is stored in the cloud server, and when it is necessary to detect the power semiconductor device to be tested by using the artificial intelligence model, the cloud server detects the power semiconductor device to be tested by using the artificial intelligence model, and obtains a detection result.

A specific scheme for the cloud server to detect the power semiconductor device to be tested by using the artificial intelligence model in this example will be described in detail below with reference to the accompanying drawings.

Referring to fig. 5, the figure is a schematic diagram of a test system for a power semiconductor device according to an embodiment of the present application.

The cloud server 100 is configured to input data of the power semiconductor device to be tested, which is obtained from the testing machine, into the pre-trained artificial intelligence model, and output of the pre-trained artificial intelligence model is a diagnosis result of the power semiconductor device to be tested. Then, the cloud server 100 transmits the diagnosis result to the test machine 200, and the test machine 200 finally outputs the diagnosis result.

It is understood that, in the solution provided in the embodiment of the present application, the cloud server 100 stores the artificial intelligence model trained by the cloud server. When the artificial intelligence model is used to detect the power semiconductor devices to be tested, the cloud server 100 may obtain data of the power semiconductor devices to be tested from the testing machine 200, and input the data into the artificial intelligence model trained in advance to obtain a diagnosis result. Then, the cloud server 100 sends the obtained diagnosis result to the test machine 200 again, so that the test machine 200 at the end board outputs the diagnosis result, that is, the test machine 200 receives only the diagnosis result and does not test the power semiconductor device by itself.

As another possible implementation, the step of inputting the data of the power semiconductor to be tested into the pre-selected trained artificial intelligence model and obtaining the diagnosis result in the embodiment of the present application may also be performed by a testing machine at the board end. The following first describes the training process of the AI model.

Referring to fig. 6, the figure is a schematic diagram of another power semiconductor device test system according to an embodiment of the present application.

The tester 200 is also configured to collect history data of the plurality of power semiconductor devices, and transmit the collected history data of the plurality of power semiconductor devices to the cloud server 100, so that the cloud server 100 can train the artificial intelligence model through the historical data of the plurality of power semiconductor devices. The history data includes at least one of chip test data of the plurality of power semiconductor devices or module function test data after packaging in the above-described embodiment. It should be noted that, after the obtained artificial intelligence model is trained in the embodiment of the present application, the cloud server 100 sends the obtained artificial intelligence model to the test machine 200 at the board end, and the test machine 200 stores the artificial intelligence model.

A specific scheme of the testing machine for testing the power semiconductor device to be tested by using the artificial intelligence model in this example will be specifically described below with reference to the accompanying drawings.

Referring to fig. 7, the figure is a schematic diagram of a test system for a power semiconductor device according to an embodiment of the present application.

Cloud server 100 is configured to send the pre-trained artificial intelligence model to test machine 200. The tester 200 is used for inputting data of the power semiconductor device to be tested into a pre-trained artificial intelligence model, and the output of the pre-trained artificial intelligence model is a diagnosis result of the power semiconductor device to be tested.

It is understood that, in the solution provided in this embodiment of the present application, the test machine 200 will receive the artificial intelligence model sent by the cloud server 100, and store the artificial intelligence model. When the artificial intelligence model is required to be used for detecting the power semiconductor device to be tested, the testing machine 200 inputs the data of the power semiconductor device to be tested into the artificial intelligence model, and obtains and outputs a diagnosis result.

It should be noted that, in the embodiment of the present application, because the artificial intelligence model is stored in the testing machine, the testing machine can directly utilize the artificial intelligence model, and the single machine completes the task of detecting the power semiconductor device to be tested, and does not need to communicate with the cloud server at the cloud end, so that when the network between the testing machine at the board end and the cloud server at the cloud end is interrupted, the testing system of the power semiconductor device in the embodiment of the present application cannot obtain the diagnosis result of the power semiconductor device to be tested. Therefore, the test system for the power semiconductor device provided by the embodiment of the application can reduce the network dependence between the cloud server and the tester, and improve the real-time performance of the test system for the power semiconductor device for obtaining the diagnosis result in the embodiment of the application.

In summary, in the test system provided in the embodiment of the present application, the cloud server at the cloud end may input the data of the power semiconductor to be tested into the artificial intelligence model of the preselection training and obtain the diagnosis result, or the test machine at the board end may input the data of the power semiconductor to be tested into the artificial intelligence model of the preselection training and obtain the diagnosis result. When the cloud server obtains the diagnosis result, the cloud server can directly store the trained artificial intelligence model without issuing the model to a testing machine at the board end. When the board-end tester obtains the diagnosis result, the cloud server needs to issue the pre-trained artificial intelligence model to the tester, and the tester can complete the test of the power semiconductor device to be tested by a single machine. When the test machine obtains the diagnosis result, the cloud server does not need to issue the diagnosis result to the test machine, so that the test machine obtains the diagnosis result without being influenced by network faults and has low dependence on the network.

The test system provided by the embodiment of the application can comprise one test machine or a plurality of test machines. The working principle of a test system including multiple test machines provided by the embodiment of the present application is described below with reference to the accompanying drawings.

Referring to fig. 8, a schematic diagram of a test system for a power semiconductor device including multiple testers according to an embodiment of the present disclosure is shown.

The test system for the power semiconductor device provided by the embodiment of the application at least comprises the following two test machines: a first tester 201 and a second tester 202;

the cloud server 100 is specifically configured to train according to first history data sent by the first testing machine 201 and second history data sent by the second testing machine 202 to obtain a global artificial intelligence model, test the power semiconductor device to be tested corresponding to the first testing machine 201 by using the global artificial intelligence model to obtain a first diagnostic result, test the power semiconductor device to be tested corresponding to the second testing machine 202 by using the global artificial intelligence model to obtain a second diagnostic result, send the first diagnostic result to the first testing machine 201, and send the second diagnostic result to the second testing machine 202.

It can be understood that the cloud server provided by the embodiment of the application utilizes the first historical data and the second historical data to train the global artificial intelligence model, and can make full use of all historical data, so that the generalization capability of the global artificial intelligence model is improved, the global artificial intelligence model with higher accuracy is utilized to diagnose the power semiconductor device to be tested, the universality of the artificial intelligence model can be improved, and the accuracy of new device diagnosis is increased.

The cloud server provided by the embodiment of the application can directly send the diagnosis result to the testing machine at the board end, and can also send the trained global artificial intelligence model to the testing machine to complete the testing by each testing machine, and the embodiment of the application is not limited herein.

Referring to fig. 9, a schematic diagram of another testing system for a power semiconductor device including multiple testers according to an embodiment of the present disclosure is shown.

The test system for the power semiconductor device provided by the embodiment of the application at least comprises the following two test machines: a first testing machine 201 and a second testing machine 202.

The cloud server 100 is specifically configured to perform training according to first history data sent by the first testing machine 201 and second history data sent by the second testing machine 202 to obtain a global artificial intelligence model, and send the global artificial intelligence model to the first testing machine 201 and the second testing machine 202.

The first tester 201 is used for testing the corresponding power semiconductor device to be tested by using the global artificial intelligence model; the second testing machine 202 is provided with a second testing machine, the method is used for testing the corresponding power semiconductor device to be tested by utilizing the global artificial intelligence model.

According to the test system provided by the embodiment of the application, the cloud server directly issues the global artificial intelligence model, the test machine can directly utilize the stored global artificial intelligence model, the single machine completes the task of detecting the power semiconductor device to be tested, and communication with the cloud server at the cloud end is not needed. The power semiconductor device testing system and the testing method avoid the situation that the testing system of the power semiconductor device cannot obtain the diagnosis result of the power semiconductor device to be tested when the network between the testing machine at the board end and the cloud server at the cloud end is interrupted. Therefore, the test system for the power semiconductor device provided by the embodiment of the application can reduce the dependence between the cloud server and the tester on the network stability, and improve the real-time performance of the test system for the power semiconductor device for obtaining the diagnosis result in the embodiment of the application.

As can be seen from the above, the test system for a power semiconductor device provided in the embodiment of the present application can obtain the global artificial intelligence model through all the historical data, so as to obtain the detection result by using the global artificial intelligence model. However, considering that there may be some systematic errors in the testing machines, the systematic errors in the measured data may be different between different testing machines, and such errors may be related to the testing machines themselves. Therefore, the embodiment of the application also provides a test system of the power semiconductor device, and after the global artificial intelligence model is obtained, the system finely adjusts the global artificial intelligence model according to historical data obtained by different test machines to obtain local artificial intelligence models corresponding to different test machines.

Referring to fig. 10, a schematic diagram of a testing system for a power semiconductor device including multiple testers according to an embodiment of the present disclosure is shown.

As shown in fig. 10, the test system for a power semiconductor device provided in the present application includes at least the following two test machines: a first testing machine 201 and a second testing machine 202.

The cloud server 100 is specifically configured to train according to first history data sent by the first testing machine 201 and second history data sent by the second testing machine 202 to obtain a global artificial intelligence model, adjust the global artificial intelligence model by using the first history data to obtain a first artificial intelligence model, and adjust the global artificial intelligence model by using the second history data to obtain a second artificial intelligence model; the first artificial intelligence model is used for testing the power semiconductor device to be tested corresponding to the first testing machine 201 to obtain a first diagnosis result, the second artificial intelligence model is used for testing the power semiconductor device to be tested corresponding to the second testing machine 202 to obtain a second diagnosis result, the first diagnosis result is sent to the first testing machine 201, and the second diagnosis result is sent to the second testing machine 202.

In the embodiment of the present application, the first history data sent by the first test machine 201 is history data corresponding to the first test machine 201. That is, the first history data includes data measured directly or indirectly by the first testing machine 201. The second history data sent by first test machine 202 is history data corresponding to second test machine 202.

According to the embodiment of the application, the first historical data and the second historical data are used for training the global artificial intelligence model, all historical data can be fully utilized, and the test accuracy of the global artificial intelligence model is improved. On the basis, considering that the test performances of different test machines may have differences, the application also adjusts the global artificial intelligence model by using the first historical data to obtain the first artificial intelligence model, and adjusts the global artificial intelligence model by using the second historical data to obtain the second artificial intelligence model, so that the local artificial intelligence model (the first artificial intelligence model) aiming at the characteristics of the first test machine and the local artificial intelligence model (the second artificial intelligence model) aiming at the characteristics of the second test machine can be obtained, and the test accuracy of the artificial intelligence model used in the actual test process is further improved.

For example, the test system for a power semiconductor device provided in the embodiments of the present application includes: the test machine Q sends the history data of the test machine Q to the cloud server, and the test machine W sends the history data of the test machine W to the cloud server. The cloud server trains out a global artificial intelligence model using the historical data of tester Q and the historical data of tester W together. Then, the cloud server adjusts the global artificial intelligence model by using the historical data of the tester Q to obtain a local artificial intelligence model corresponding to the tester Q, and obtains the power semiconductor device to be tested corresponding to the tester Q by using the local artificial intelligence model to test. And the cloud server adjusts the global artificial intelligence model by using the historical data of the test machine W to obtain a local artificial intelligence model corresponding to the test machine W, and tests the power semiconductor device to be tested corresponding to the test machine W by using the local artificial intelligence model.

The cloud server provided by the embodiment of the application can directly send the diagnosis result to the board-end testing machine, and can also send the trained artificial intelligence model to the board-end testing machine, and the board-end testing machine tests the power semiconductor device by using the trained artificial intelligence model to obtain the diagnosis result, which is described in detail below with reference to the attached drawings.

Referring to fig. 11, a schematic diagram of a testing system for a power semiconductor device including multiple testers according to an embodiment of the present disclosure is shown.

The test system for the power semiconductor device provided by the application at least comprises the following two test machines: a first tester 201 and a second tester 202;

the cloud server 100 is specifically configured to train according to first history data sent by the first testing machine 201 and second history data sent by the second testing machine 202 to obtain a global artificial intelligence model, adjust the global artificial intelligence model by using the first history data to obtain a first artificial intelligence model, send the first artificial intelligence model to the first testing machine 201, and adjust the global artificial intelligence model by using the second history data to obtain a second artificial intelligence model, and send the second artificial intelligence model to the second testing machine 202.

The first testing machine 201 is specifically configured to test a corresponding power semiconductor device to be tested by using a first artificial intelligence model; the second testing machine 202 is specifically configured to test the corresponding power semiconductor device to be tested by using the second artificial intelligence model.

In the embodiment of the application, the cloud server directly issues the first artificial intelligence model and the second artificial intelligence model to the first testing machine and the second testing machine respectively, the testing machines can directly utilize the corresponding artificial intelligence models, the single machine completes the task of detecting the power semiconductor devices to be tested, communication with the cloud server at the cloud end is not needed, and when network interruption between the testing machine at the board end and the cloud server at the cloud end is avoided, the testing system of the power semiconductor devices in the embodiment of the application cannot obtain the diagnosis results of the power semiconductor devices to be tested. Therefore, the test system of the power semiconductor device provided by the embodiment of the application can reduce the dependence on network stability between the cloud server and the test machine, and improve the real-time performance of the test system of the power semiconductor device for obtaining the diagnosis result.

As can be seen from the above, the test system for a power semiconductor device provided in the embodiment of the present application may directly use all historical data to obtain the global artificial intelligence model and directly use the global artificial intelligence model, or may use historical data corresponding to different testing machines after obtaining the global artificial intelligence model to adjust the global artificial intelligence model to obtain the local artificial intelligence model, and use the local artificial intelligence model to detect the power semiconductor device of the corresponding testing machine.

In order to obtain a higher-precision artificial intelligence model, as a possible implementation manner, the embodiment of the application can also perform subsequent tests on the power semiconductor device diagnosed as invalid to verify whether the power semiconductor device is invalid or not, and further verify whether the AI model is accurate or not, and if the power semiconductor device is verified to be invalid, the AI model is inaccurate and needs to be adjusted. And continuously collecting the operation data of the power semiconductor device which is judged to be not failed at the application side, uploading the data of the two power semiconductor devices to a cloud server, and using the two data by the cloud server to adjust the AI model.

In order to make the accuracy of the artificial intelligence model higher in the embodiment of the present application, as a possible implementation manner, the testing machine in the embodiment of the present application is further configured to fine-tune the pre-trained model in combination with data of the testing machine side, diagnose the power semiconductor device to be tested by using the fine-tuned model, and output a diagnosis result. It should be noted that the fine tuning in the embodiment of the present application may include parameter optimization on a pre-trained model. As one example, a method of transfer learning may be utilized to fine-tune a pre-trained AI model. For example, the trimmed AI model may be used to test whether the wafer has failed.

Considering that a small amount of historical data collected by the tester is stored in the tester in the embodiment of the present application, but the data is not uploaded to the cloud server, as a possible implementation manner, the data on the tester side in the embodiment of the present application may further include data that is not uploaded to the cloud server. It can be understood that the testing machine in the embodiment of the present application generally uploads the data collected by the testing machine to the cloud server periodically, and the testing machine in the embodiment of the present application generally stores a certain amount of historical data which is not uploaded, so that the testing machine can perform fine adjustment on the AI model by using the data which is not uploaded, and then detect the power semiconductor device based on the fine-adjusted AI model. After the AI model is finely adjusted by the testing machine in the embodiment of the application, the finely adjusted AI model can be uploaded to a cloud server, so that the AI model can be uniformly managed by the cloud server.

In summary, the test system for the power semiconductor device provided in the embodiment of the present application, in consideration of the limitation of the computing capability and the storage capability of the test machine, trains the artificial intelligence model by using the cloud server with relatively large computing capability and storage capability, directly obtains a diagnosis result by using the artificial intelligence model in the cloud server at the cloud end, and also sends the artificial intelligence model to the test machine at the board end, and obtains a diagnosis result by using the artificial intelligence model by using the test machine, and also can send only the artificial intelligence model which is not completely trained to the test machine at the board end, and the test machine at the board end completes the training of the artificial intelligence model according to the data stored in the test machine, and obtains a diagnosis result by using the artificial intelligence model which is completed by training.

Server embodiment

According to the test system of the power semiconductor device provided by the embodiment, the embodiment of the application further provides the cloud server.

Referring to fig. 12, the figure is a schematic structural diagram of a cloud server provided in an embodiment of the present application.

As shown in fig. 12, a cloud server provided in an embodiment of the present application includes: a first transceiving device 1201 and a first controller 1202.

The first transceiving equipment 1201 is configured to receive test data of the power semiconductor device to be tested, which is sent by the tester.

A first controller 1202, configured to train in advance with historical data of a plurality of power semiconductor devices to obtain an artificial intelligence model; the system is also used for inputting the data of the power semiconductor device to be tested into the artificial intelligence model, and the output of the pre-trained AI model is the diagnosis result of the power semiconductor device to be tested; the first transceiver is also used for sending the diagnosis result to the tester;

or, the first controller 1202 is configured to send the pre-trained artificial intelligence model to the tester, so that the tester diagnoses the power semiconductor device to be tested by using the pre-trained artificial intelligence model.

The specific implementation manner of the cloud server provided in the embodiment of the present application may refer to the description of the above test system embodiment on the cloud server, and is only briefly described here.

In a possible implementation manner of the embodiment of the application, the first controller is specifically configured to, when an unsupervised learning model is used for training and obtaining an artificial intelligence model, specifically obtain failure types of a plurality of power semiconductor devices, extract all test items corresponding to each failure type in test items of the power semiconductor devices by using expert knowledge, form a test item subset by all test items corresponding to each failure type, perform abnormal level detection on each test item subset by using the unsupervised learning model, obtain a first score and a second score for an abnormal device and a normal device, and obtain a sum of the first score and the second score for all test item subsets of each power semiconductor device as an abnormal level total score; and when the total abnormal level score is larger than a preset score threshold value, judging that the power semiconductor device is an abnormal device.

In a possible implementation manner of the embodiment of the present application, the cloud server corresponds to at least two of the following test machines: a first testing machine and a second testing machine. The first controller is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, adjusting the global artificial intelligence model by using the first historical data to obtain a first artificial intelligence model, and adjusting the global artificial intelligence model by using the second historical data to obtain a second artificial intelligence model; testing the power semiconductor device to be tested corresponding to the first testing machine by using the first artificial intelligent model to obtain a first diagnosis result, and testing the power semiconductor device to be tested corresponding to the second testing machine by using the second artificial intelligent model to obtain a second diagnosis result; the first transceiver device is specifically configured to send the first diagnostic result to the first testing machine, and is further configured to send the second diagnostic result to the second testing machine.

In a possible implementation manner of the embodiment of the present application, the cloud server corresponds to at least two of the following test machines: a first testing machine and a second testing machine. The first controller is specifically used for training according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, testing a power semiconductor device to be tested corresponding to the first testing machine by using the global artificial intelligence model to obtain a first diagnosis result, and testing the power semiconductor device to be tested corresponding to the second testing machine by using the global artificial intelligence model to obtain a second diagnosis result; the first transceiver is specifically configured to send the first diagnostic result to the first tester, and is further configured to send the second diagnostic result to the second tester.

In a possible implementation manner of the embodiment of the present application, the first transceiver device is further configured to receive update data sent by the tester; the first controller is also used for updating the artificial intelligence model according to the updating data.

The cloud server provided by the embodiment of the application may further include a first storage device (not shown in the figure) besides the first transceiver device and the first controller, where the first storage device may be used to store history data and update data uploaded by the tester, so as to facilitate subsequent processing by the first controller. Before the first controller utilizes the historical data to train the AI model, in order to make the parameters of the trained AI model more accurate, some interference data can be eliminated, and data cleaning can be specifically carried out.

In summary, the cloud server in the embodiment of the application can train the artificial intelligence model by using the cloud server with strong computing capability and storage capability, so that the artificial intelligence model with high diagnosis accuracy is obtained. After the artificial intelligence model is obtained, the artificial intelligence model can be directly used in the cloud server to obtain the diagnosis result, and the artificial intelligence model can be only provided for the testing machine, but the direct diagnosis result is not provided.

Test machine embodiment

Based on the test system for the power semiconductor device and the cloud server provided by the embodiments, the embodiments of the present application further provide a testing machine, which is described in detail below with reference to the accompanying drawings.

Referring to fig. 13, the drawing is a schematic structural diagram of a testing machine according to an embodiment of the present disclosure.

As shown in fig. 13, the testing machine in the embodiment of the present application includes: a second transceiving device 1301 and a second controller 1302;

the second transceiver 1301 is configured to receive an artificial intelligence model sent by the cloud server, where the artificial intelligence model is obtained by the cloud server through training in advance by using historical data of the plurality of power semiconductor devices;

a second controller 1302 for obtaining data of the power semiconductor device to be tested; and inputting the data of the power semiconductor device to be tested into a pre-trained artificial intelligence model, wherein the output of the pre-trained AI model is the diagnosis result of the power semiconductor device to be tested.

The test machine provided in the embodiment of the present application generally refers to any one test machine, one cloud server may correspond to multiple test machines provided in the embodiment of the present application, and the specific working manner of the test machine may refer to the introduction of the embodiment of the test system to the test machine, which is only briefly described here.

In a possible implementation manner of the embodiment of the present application, the second controller is further configured to collect historical data of the plurality of power semiconductor devices, and send the collected historical data of the plurality of power semiconductor devices to the cloud server; the historical data includes at least one of chip test data of the plurality of power semiconductor devices or module function test data after packaging.

In a possible implementation manner of the embodiment of the application, the second controller is further configured to compare data of the power semiconductor device to be tested with preset upper and lower limit values, and determine that the power semiconductor device to be tested is an abnormal device when the data of the power semiconductor device to be tested exceeds the upper and lower limit values. In the application, the test machine can directly obtain the diagnosis result sent by the cloud server, also can obtain the artificial intelligence model sent by the cloud server, and obtains the diagnosis result in the test machine through the artificial intelligence model.

In summary, the testing machine provided by the embodiment of the application has limited computing capability and storage capability, and therefore, the artificial intelligence model with high diagnosis accuracy is obtained by sending data to the cloud server and training the artificial intelligence model by using the cloud server with strong computing capability and storage capability of the cloud server.

It should be understood that, in this application, "at least one" means one or more, "a plurality" means two or more. "and/or" for describing an association relationship of associated objects, indicating that there may be three relationships, e.g., "a and/or B" may indicate: only A, only B and both A and B are present, wherein A and B may be singular or plural. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one of the following" or similar expressions refer to any combination of these items, including any combination of the singular or plural items. For example, at least one (one) of a, b, or c, may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.

Claims

1. A test system for a power semiconductor device, comprising: the system comprises a cloud server and at least one testing machine;

the cloud server is used for training by utilizing historical data of a plurality of power semiconductor devices in advance to obtain an artificial intelligence model;

the tester is used for obtaining data of the power semiconductor device to be tested;

the cloud server or the tester is used for inputting the data of the power semiconductor device to be tested into the artificial intelligence model, and the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested.

2. The system of claim 1, wherein the cloud server is configured to input data of the power semiconductor devices to be tested into the artificial intelligence model, and an output of the artificial intelligence model is a diagnosis result of the power semiconductor devices to be tested, and the diagnosis result is sent to the tester.

3. The system of claim 1, wherein the cloud server is configured to send a pre-trained artificial intelligence model to the testing machine;

the tester is used for inputting the data of the power semiconductor device to be tested into the artificial intelligence model, and the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested.

4. The system of any one of claims 1-3, wherein the tester is further configured to collect historical data of the plurality of power semiconductor devices and send the collected historical data of the plurality of power semiconductor devices to the cloud server; the historical data comprises at least one item of chip test data of the plurality of power semiconductor devices, packaged module function test data or related data when actual application is abnormal;

the cloud server is further used for training by using at least one of a supervised learning model or an unsupervised learning model by using historical data of the plurality of power semiconductor devices to obtain the artificial intelligence model.

5. The system according to any one of claims 1 to 4, wherein the cloud server obtains the artificial intelligence model by using the unsupervised learning model training, specifically to obtain failure types of the plurality of power semiconductor devices, extracts all test items corresponding to each failure type from the test items of the power semiconductor devices by using expert knowledge, forms a test item subset by using all test items corresponding to each failure type, distinguishes an abnormal subset and a normal subset by using the unsupervised learning model for each test item subset, obtains a first score and a second score for the abnormal subset and the normal subset respectively, and obtains a sum of the first score and the second score for all test item subsets of each power semiconductor device as an abnormal level total score; and when the total abnormal level score is greater than or equal to a preset score threshold value, judging that the power semiconductor device is an abnormal device.

6. The system of claim 5, wherein the cloud server is further configured to use expert knowledge to score different first scores for anomalous devices in different subsets of test items.

7. The system according to any one of claims 1 to 4, wherein the cloud server obtains the artificial intelligence model by using the supervised learning model training, and is specifically configured to obtain relevant data of the power semiconductor device in an abnormal state as a data tag, extract the data feature of the data tag by using the supervised learning model, and obtain an abnormal level total score of each power semiconductor device by using the data feature, and when the abnormal level total score is greater than a preset score threshold, determine that the power semiconductor device is an abnormal device.

8. The system according to any one of claims 1 to 7, wherein the tester is further configured to compare data of the power semiconductor device to be tested with preset upper and lower limit values, and determine that the power semiconductor device to be tested is an abnormal device when the data of the power semiconductor device to be tested exceeds the upper and lower limit values.

9. The system of claim 2, wherein the system comprises at least two of the following testing machines: a first tester and a second tester;

the cloud server is specifically configured to train according to first history data sent by the first test machine and second history data sent by the second test machine to obtain a global artificial intelligence model, adjust the global artificial intelligence model by using the first history data to obtain a first artificial intelligence model, and adjust the global artificial intelligence model by using the second history data to obtain a second artificial intelligence model; and testing the power semiconductor device to be tested corresponding to the first testing machine by using the first artificial intelligent model to obtain a first diagnosis result, testing the power semiconductor device to be tested corresponding to the second testing machine by using the second artificial intelligent model to obtain a second diagnosis result, sending the first diagnosis result to the first testing machine, and sending the second diagnosis result to the second testing machine.

10. The system of claim 2, wherein the system comprises at least two of the following testing machines: a first tester and a second tester;

the cloud server is specifically configured to train according to first history data sent by the first testing machine and second history data sent by the second testing machine to obtain a global artificial intelligence model, test the power semiconductor device to be tested corresponding to the first testing machine by using the global artificial intelligence model to obtain a first diagnostic result, test the power semiconductor device to be tested corresponding to the second testing machine by using the global artificial intelligence model to obtain a second diagnostic result, send the first diagnostic result to the first testing machine, and send the second diagnostic result to the second testing machine.

11. The system of claim 3, wherein the system comprises at least two of the following testing machines: a first tester and a second tester;

the cloud server is specifically configured to train according to first history data sent by the first test machine and second history data sent by the second test machine to obtain a global artificial intelligence model, adjust the global artificial intelligence model by using the first history data to obtain a first artificial intelligence model, send the first artificial intelligence model to the first test machine, adjust the global artificial intelligence model by using the second history data to obtain a second artificial intelligence model, and send the second artificial intelligence model to the second test machine;

the first tester is specifically used for diagnosing the corresponding power semiconductor device to be tested by using the first artificial intelligent model;

the second testing machine is specifically configured to diagnose the corresponding power semiconductor device to be tested by using the second artificial intelligence model.

12. The system of claim 3, wherein the system comprises at least two of the following testing machines: a first tester and a second tester;

the cloud server is specifically configured to train according to first history data sent by the first testing machine and second history data sent by the second testing machine to obtain a global artificial intelligence model, and send the global artificial intelligence model to the first testing machine and the second testing machine;

the first testing machine is used for diagnosing the corresponding power semiconductor device to be tested by utilizing the global artificial intelligence model;

and the second tester is used for diagnosing the corresponding power semiconductor device to be tested by utilizing the global artificial intelligence model.

13. The system of any one of claims 1-12, wherein the tester is further configured to send update data to the cloud server;

the cloud server is further used for updating the artificial intelligence model according to the updating data.

14. The system according to any one of claims 3 to 13, wherein the tester is further configured to fine-tune the pre-trained AI model in combination with tester-side data, diagnose the power semiconductor device to be tested using the fine-tuned model, and output a diagnosis result.

15. A cloud server, comprising: a first transceiver device and a first controller;

the first transceiver is used for receiving test data of the power semiconductor device to be tested, which is sent by the tester;

the first controller is used for training by utilizing historical data of a plurality of power semiconductor devices in advance to obtain an artificial intelligence model; the system is also used for inputting the data of the power semiconductor device to be tested into the artificial intelligence model, and the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested; the first transceiver device is further configured to send the diagnosis result to the tester;

or the like, or, alternatively,

the first controller is used for sending the artificial intelligence model to the testing machine so that the testing machine diagnoses the power semiconductor device to be tested by using the artificial intelligence model.

16. The cloud server according to claim 15, wherein the first controller is specifically configured to, when the artificial intelligence model is obtained through unsupervised learning model training, specifically, obtain failure types of the plurality of power semiconductor devices, extract all test items corresponding to each failure type in the test items of the power semiconductor devices by using expert knowledge, form a test item subset by using all test items corresponding to each failure type, perform anomaly level detection on each test item subset by using the unsupervised learning model, obtain a first score and a second score for each abnormal device and each normal device, and obtain a sum of the first score and the second score for each test item subset of the power semiconductor devices as an anomaly level total score; and when the total abnormal level score is greater than a preset score threshold value, judging that the power semiconductor device is an abnormal device.

17. The cloud server of claim 15 or 16, wherein the cloud server corresponds to at least two of the following testing machines: a first tester and a second tester;

the first controller is specifically configured to train according to first historical data sent by the first testing machine and second historical data sent by the second testing machine to obtain a global artificial intelligence model, adjust the global artificial intelligence model by using the first historical data to obtain a first artificial intelligence model, and adjust the global artificial intelligence model by using the second historical data to obtain a second artificial intelligence model; testing the power semiconductor device to be tested corresponding to the first testing machine by using the first artificial intelligent model to obtain a first diagnosis result, and testing the power semiconductor device to be tested corresponding to the second testing machine by using the second artificial intelligent model to obtain a second diagnosis result;

the first transceiver device is specifically configured to send the first diagnostic result to the first tester, and is further configured to send the second diagnostic result to the second tester.

18. The cloud server of claim 15 or 16, wherein the cloud server corresponds to at least two of the following testing machines: a first tester and a second tester;

the first controller is specifically configured to train according to first history data sent by the first testing machine and second history data sent by the second testing machine to obtain a global artificial intelligence model, test the power semiconductor device to be tested corresponding to the first testing machine by using the global artificial intelligence model to obtain a first diagnostic result, and test the power semiconductor device to be tested corresponding to the second testing machine by using the global artificial intelligence model to obtain a second diagnostic result;

19. The cloud server of any of claims 14-17, wherein the first transceiver device is further configured to receive update data sent by the tester;

the first controller is further configured to update the artificial intelligence model according to the update data.

20. A testing machine, comprising: a second transceiver device and a second controller;

the second transceiver is used for receiving an artificial intelligence model sent by the cloud server, and the artificial intelligence model is obtained by training the cloud server in advance by using historical data of a plurality of power semiconductor devices;

the second controller is used for obtaining data of the power semiconductor device to be tested; and inputting the data of the power semiconductor device to be tested into the artificial intelligence model, wherein the output of the artificial intelligence model is the diagnosis result of the power semiconductor device to be tested.

21. The testing machine of claim 20, wherein the second controller is further configured to collect historical data of the plurality of power semiconductor devices and send the collected historical data of the plurality of power semiconductor devices to the cloud server; the historical data includes at least one of chip test data or packaged module function test data of the plurality of power semiconductor devices.

22. The testing machine of claim 20 or 21, wherein the second controller is further configured to compare data of the power semiconductor device to be tested with preset upper and lower limit values, and determine that the power semiconductor device to be tested is an abnormal device when the data of the power semiconductor device to be tested exceeds the upper and lower limit values.