Surface temperature detection method for chip-level GaAs power device

Abstract

The invention discloses a surface temperature detection method for a chip-level GaAs power device, and belongs to the technical field of semiconductor device temperature detection. The method comprises the following steps: acquiring a surface temperature image of a device in a working state, and identifying a plurality of virtual heat source points of the device according to the surface temperature image; arranging a plurality of thermocouples on the surface of the device; acquiring temperature images of the surface of the device at different environment temperatures, calculating a first test temperature of the virtual heat source point according to the temperature images, and acquiring a second test temperature of the virtual heat source point through a thermocouple; fitting a temperature compensation model of each virtual heat source point according to the environment surface temperature difference value and the test temperature difference value; and acquiring the environment temperature and the working temperature data of the virtual heat source point in real time, and correcting the working temperature data according to the environment temperature and the temperature compensation model. The working temperature data acquired by the thermocouple are compensated in real time through the environment temperature and the temperature compensation model, and high-precision device surface temperature detection is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of semiconductor device temperature detection, in particular to a surface temperature detection method for a chip-level GaAs power device. Background Art

[0002] Power devices are widely used in various electronic devices. Their performance (such as on-resistance, current capacity, switching speed, etc.) and operating environment (such as operating temperature, voltage, current maximum value, etc.) are important factors that determine their final function and performance. Since semiconductor power devices generate a lot of heat when working, good thermal design and thermal management are required to ensure their normal and stable operation and increase their service life.

[0003] Among them, GaAs semiconductors have higher electron mobility and saturated electron flow speed, making chip-level GaAs power devices superior to many other types of semiconductor power devices, such as silicon semiconductor power devices, in terms of operating speed and frequency. Although GaAs power devices have excellent working performance, their energy conversion requirements will cause greater heat generation, so accurate temperature detection is required to provide data support for the thermal management of power devices.

[0004] On the one hand, using thermocouples to measure the surface temperature of semiconductor power devices is one of the commonly used temperature measurement methods. However, since thermocouples are contact detection methods, this method is prone to large detection errors. At the same time, thermocouples measure the single-point temperature on the surface of the device, and their fixing method and position as well as the ambient temperature will affect the measurement accuracy. Therefore, using thermocouples to measure the temperature of semiconductor power devices requires high-precision calibration to ensure the accuracy of the measurement results.

[0005] On the other hand, infrared thermal imagers can determine the surface temperature of an object without contact by capturing infrared photons emitted or reflected by power devices. The advantage of thermal imagers is that they can accurately measure the surface temperature of power devices and obtain the overall temperature distribution of power devices. This is of great value in understanding the thermal behavior of devices, determining possible thermal problem areas, and evaluating heat dissipation designs. However, the high cost of infrared thermal imagers prevents this method from being widely used in actual test scenarios where a large number of power devices need to be continuously tested. Summary of the invention

[0006] In order to solve the problems existing in the above background technology, the present invention adopts the following technical solutions:

[0007] The present application provides a surface temperature detection method for a chip-level GaAs power device, comprising the steps of:

[0008] Acquire a surface temperature image of the device in a working state, and identify several virtual heat source points of the device according to the surface temperature image;

[0009] A plurality of thermocouples are arranged on the surface of the device, and the thermocouples are used to collect the temperature of the virtual heat source points; the number of the thermocouples is the same as the number of the virtual heat source points;

[0010] Acquire a temperature image of the device surface at different ambient temperatures, calculate a first test temperature of a virtual heat source point according to the temperature image, and simultaneously acquire a second test temperature of the virtual heat source point through a thermocouple;

[0011] Calculating a test temperature difference according to the first test temperature and the second test temperature, calculating an ambient surface temperature difference according to the second test temperature and the ambient temperature, and fitting a temperature compensation model for each virtual heat source point according to the ambient surface temperature difference and the test temperature difference;

[0012] The ambient temperature and the operating temperature data of the virtual heat source point are acquired in real time, and the operating temperature data are corrected according to the ambient temperature and the temperature compensation model.

[0013] As a preferred solution of the present application, the method of identifying several virtual heat source points of a device according to a surface temperature image specifically includes the steps of:

[0014] Convert the surface temperature image from a color image to a grayscale image;

[0015] Set the reference temperature, and draw several closed isotherms in the grayscale image according to the grayscale value corresponding to the reference temperature;

[0016] The point with the lowest gray value within the closed isotherm is recorded as the virtual heat source point.

[0017] As a preferred embodiment of the present application, the reference temperature is expressed as:

[0018] T ref = k(T j,max -P×R jc ),

[0019] Among them, T j,max is the standard maximum junction temperature, R jc is the standard thermal resistance; P is the working power of the device; k is the proportional adjustment coefficient.

[0020] As a preferred solution of the present application, the method of identifying several virtual heat source points of the device according to the surface temperature image further includes the steps of:

[0021] The real heat source point in each closed isotherm is identified according to the temperature distribution gradient characteristics; when the virtual heat source point does not overlap with the real heat source point and the virtual heat source point is on the line connecting the first real heat source point and the second real heat source point, the set virtual heat source point is deleted and the first real heat source point and the second real heat source point are respectively set as virtual heat source points; wherein the first real heat source point is the point with the lowest grayscale value among the real heat source points of the closed isotherm; and the second real heat source point is the point with the second lowest grayscale value among the real heat source points of the closed isotherm.

[0022] As a preferred solution of the present application, the temperature compensation model of each virtual heat source point is fitted according to the ambient surface temperature difference and the test temperature difference, specifically comprising the steps of:

[0023] Divide the ambient surface temperature difference and the test temperature difference data into a training set and a test set;

[0024] A temperature compensation model is obtained by fitting the training set data using the least squares method; the temperature compensation model is an electrical parameter analysis model;

[0025] The temperature compensation model was tested using the test set data, and the prediction accuracy of the model was evaluated using the mean square error indicator;

[0026] Draw the error curve of the temperature compensation model, identify the overfitting or underfitting phenomenon in the training process of the temperature compensation model through the error curve, and optimize the temperature compensation model by increasing or decreasing the polynomial degree or regularization method accordingly; the error curve is used to represent the distribution of training error and verification error as the number of iterations changes.

[0027] As a preferred solution of the present application, the method further comprises the steps of:

[0028] Obtain thermally sensitive electrical parameters of power devices;

[0029] Constructing a plurality of electrical parameter analysis models, each of which corresponds to a virtual heat source point;

[0030] The electrical parameter analysis model is trained by using the thermal sensitive electrical parameters, the operating temperature data and the thermal sensitive electrical parameters;

[0031] The real-time acquired thermally sensitive electrical parameters and operating temperature data are input into the trained electrical parameter analysis model to predict the temperature of the virtual heat source point. When the ratio of the predicted result to the real-time acquired operating temperature data of the virtual heat source point is greater than a preset threshold, it is determined that the power device is faulty.

[0032] As a preferred solution of the present application, the electrical parameter analysis model is a polynomial regression model; the independent variables of the electrical parameter analysis model include thermally sensitive electrical parameters and ambient temperature; and the dependent variable of the electrical parameter analysis model is the operating temperature data of the virtual heat source point.

[0033] As a preferred solution of the present application, the temperature compensation value is expressed as:

[0034] T i,com =T i,1 -T i,2 ,

[0035] Where i is a positive integer, indicating the sequence number of the virtual heat source point; T i,com is the temperature compensation value of the ith virtual heat source point, T i,1 is the first test temperature of the ith virtual heat source point, T i,2 is the second test temperature of the i-th virtual heat source point;

[0036] The ambient surface temperature difference is expressed as:

[0037] T i,def =T i,2 -T s ,

[0038] Among them, T i,def represents the ambient surface temperature difference of the i-th virtual heat source point, T s Indicates the ambient temperature.

[0039] As a preferred solution of the present application, before converting the surface temperature image from a color image to a grayscale image, it also includes the step of equalizing the histogram of the surface temperature image to obtain a surface temperature image with optimized brightness and contrast.

[0040] As a preferred solution of the present application, before converting the surface temperature image from a color image to a grayscale image, the following steps are further included:

[0041] Select the size of the filter window, traverse each pixel in the image in turn, sort all the pixel values ​​in the filter window, and take the middle value as the new value of the pixel point; the new value of the pixel point is expressed as:

[0042] P(x,y)=med ian{Q(x+i,y+j)|i,j∈W},

[0043] Among them, P(x,y) represents the new value of the pixel, x and y represent the horizontal and vertical coordinates of the pixel respectively, Q(x+i,y+j) represents the value of the neighborhood around the corresponding pixel in the original image, W represents the filter window, i and j are the parameters that loop within the filter window, and median{} represents the operation of taking the median of the pixel values ​​in the neighborhood.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The present application detects the temperature of each virtual heat source point through a thermocouple to obtain a high-precision second test temperature. Especially for equipment that needs to work continuously for a long time, it can provide a data basis for simulating and predicting the thermal behavior of the chip under changing working conditions, so as to detect possible faults early and take necessary heat dissipation or protection measures. Furthermore, the present application collects a first temperature image through a thermal imager, calculates the test temperature difference based on the first test temperature and the second test temperature, and combines it with the ambient temperature factor to fit the temperature compensation model, so as to perform high-precision correction on the second test temperature corresponding to each virtual heat source point collected by the thermocouple; after the correction is completed, the real-time acquired ambient temperature and temperature compensation model can be used to realize real-time compensation of the working temperature data collected by the thermocouple, further realizing higher-precision device surface temperature detection.

[0046] The embodiment of the present application corrects the virtual heat source point on the line connecting the first real heat source point and the second real heat source point by identifying the real heat source point, so that the acquired virtual heat source point can accurately correspond to the first real heat source point and the second real heat source point in the closed isotherm, thereby improving the accuracy of temperature calculation, and only needs to be calculated based on the acquired working temperature data without understanding the specific physical properties or structure inside the device, so the application cost is low and the scope is wide.

[0047] The embodiment of the present application acquires the thermally sensitive electrical parameters and the environmental and operating temperature data in real time, constructs an electrical parameter analysis model based on this data, and predicts the temperature of the virtual heat source point through the model. When the deviation between the predicted temperature and the actually acquired temperature exceeds the preset threshold, it can be judged in time that the power device has a fault, thereby improving the monitoring accuracy of the power device and the timeliness of fault detection, thereby achieving the purpose of improving equipment efficiency and extending equipment service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 A schematic flow chart of a method for detecting the surface temperature of a chip-level GaAs power device according to an embodiment of the present invention;

[0050] Figure 2 It is a schematic diagram of a process of identifying several virtual heat source points of a device according to a surface temperature image according to an embodiment of the present invention;

[0051] Figure 3 It is a schematic diagram of a flow chart of fitting a temperature compensation model of each virtual heat source point according to an ambient surface temperature difference and a test temperature difference according to another embodiment of the present invention;

[0052] Figure 4 FIG. 4 is a flow chart of step S6 according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. It should be noted that relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0054] See also Figure 1 The present invention provides a surface temperature detection method for a chip-level GaAs power device, comprising the steps of:

[0055] S1. Acquire a surface temperature image of a device in a working state, and identify a number of virtual heat source points of the device according to the surface temperature image; specifically, the surface temperature image is acquired by an infrared thermal imager.

[0056] S2. Arrange a plurality of thermocouples on the surface of the device, wherein the thermocouples are used to collect the temperature of the virtual heat source points; the number of the thermocouples is the same as the number of the virtual heat source points;

[0057] S3. Obtain a temperature image of the device surface at different ambient temperatures, calculate a first test temperature of the virtual heat source point based on the temperature image, and simultaneously obtain a second test temperature of the virtual heat source point through a thermocouple; wherein the different ambient temperatures are specifically a number of arithmetic progressions within the standard working ambient temperature range; and the ambient temperature is obtained through a temperature sensor integrated with the power device.

[0058] S4, calculating the test temperature difference according to the first test temperature and the second test temperature, calculating the ambient surface temperature difference according to the second test temperature and the ambient temperature, and fitting a temperature compensation model of each virtual heat source point according to the ambient surface temperature difference and the test temperature difference;

[0059] S5. Acquire the ambient temperature and the working temperature data of the virtual heat source point in real time, and correct the working temperature data according to the ambient temperature and the temperature compensation model.

[0060] Specifically, the temperature compensation value is expressed as:

[0061] T i,com =T i,1 -T i,2 ,

[0062] Where i is a positive integer, indicating the sequence number of the virtual heat source point; T i,com is the temperature compensation value of the ith virtual heat source point, T i,1 is the first test temperature of the ith virtual heat source point, T i,2 is the second test temperature of the i-th virtual heat source point.

[0063] Specifically, the ambient surface temperature difference is expressed as:

[0064] T i,def =T i,2 -T s ,

[0065] Among them, T i,def represents the ambient surface temperature difference of the i-th virtual heat source point, T s Represents the ambient temperature. It can be understood that the difference between the surface temperature of the power device and the test temperature of its surrounding environment directly affects the heat dissipation speed of the object. Therefore, the present application can accurately reflect the heat dissipation status of the heat source by fitting the temperature compensation model through the difference between the ambient surface temperature and the test temperature difference, so that the actual temperature of the virtual heat source point of the device can be more accurately predicted through the temperature compensation model.

[0066] When the power device is in operation, there may be heat sources inside it to varying degrees. The prior art usually uses a single thermocouple to detect the surface temperature of the device, which can only obtain a single-point temperature on the surface of the device, but cannot obtain the real-time temperature of a specific high-temperature point. The present application detects the temperature of each virtual heat source point through a thermocouple to obtain a high-precision second test temperature, especially for equipment that needs to work continuously for a long time, which can provide a data basis for simulating and predicting the thermal behavior of the chip under variable working conditions, thereby discovering possible faults early and taking necessary heat dissipation or protection measures. Furthermore, the present application collects a first temperature image through a thermal imager, calculates the test temperature difference based on the first test temperature and the second test temperature, and combines it with the ambient temperature factor to fit the temperature compensation model, so as to perform high-precision correction on the second test temperature corresponding to each virtual heat source point collected by the thermocouple; after the correction is completed, the real-time acquired ambient temperature and temperature compensation model can be used to realize real-time compensation of the working temperature data collected by the thermocouple, further realizing higher-precision device surface temperature detection.

[0067] For further information, see Figure 2 The method of identifying several virtual heat source points of the device according to the surface temperature image specifically includes the steps of:

[0068] S11, converting the surface temperature image from a color image to a grayscale image. A grayscale image only contains information about grayscale changes, so grayscale images are easier to process than color images.

[0069] S12, setting a reference temperature, and drawing a number of closed isotherms in the grayscale image according to the grayscale value corresponding to the reference temperature;

[0070] S13. Record the lowest gray value point within the closed isotherm as a virtual heat source point; wherein the lowest gray value point is the highest temperature point in the region.

[0071] Specifically, the reference temperature is calculated by the standard maximum junction temperature and the standard thermal resistance; the reference temperature is expressed as:

[0072] T ref = k(T j,max -P×R jc ),

[0073] Among them, T j,max is the standard maximum junction temperature, R jc is the standard thermal resistance. The standard maximum junction temperature and standard thermal resistance are provided in the product specification. P is the working power of the device. k is the proportional adjustment coefficient. Reference temperature T refUsed to set the benchmark for heat source identification. The reference temperature is a monitoring indicator set in the embodiment of this scheme. Positions exceeding this temperature are considered possible heat sources. The proportional adjustment coefficient k is used to fine-tune the reference temperature in this embodiment. The setting of the proportional adjustment coefficient k will affect the temperature detection accuracy and the speed of finding the heat source. The specific value can be set according to the prior knowledge of the technician and experimental feedback. In one embodiment, the value range is [0.5, 0.8].

[0074] It can be understood that the virtual heat source point searched in this embodiment is not necessarily the real heat source point at the corresponding position of the working chip; for example, when an isotherm includes two working chips, and both working chips are real heat source points, the virtual heat source point searched through step S13 may be one of the two real heat source points, or it may be a point on the line connecting the two real heat source points. The specific position of the virtual heat source point is determined by the structure of the power device and the performance of the working chip. Therefore, in this embodiment, the smaller the coefficient k is, the farther the temperature corresponding to the drawn isotherm is from the highest junction temperature, so the real heat source points with lower temperatures in the device can be more accurately included in the isotherm. However, setting the proportional adjustment coefficient k value to a smaller value also increases the area for searching the virtual heat source point, which may lead to the problem of including more than one real heat source point in the isotherm, thereby reducing the accuracy of the test result when the performance of one of the real heat source points deviates from the standard. To solve this problem, as a preferred embodiment, the method of identifying several virtual heat source points of the device based on the surface temperature image also includes the steps of:

[0075] S14. Identify the real heat source point in each closed isotherm according to the temperature distribution gradient characteristics; when the virtual heat source point does not overlap with the real heat source point and the virtual heat source point is on the line connecting the first real heat source point and the second real heat source point, delete the set virtual heat source point and set the first real heat source point and the second real heat source point as virtual heat source points respectively.

[0076] Among them, the first real heat source point is the point with the lowest gray value among the real heat source points of the closed isotherm; the second real heat source point is the point with the second lowest gray value among the real heat source points of the closed isotherm. The temperature distribution gradient represents the rate of change of temperature in space. The temperature distribution gradient is obtained by calculating the temperature difference between each point and its surrounding points. These temperature gradient values ​​are sorted, and the point with the largest temperature gradient value (i.e., the most drastic temperature change) is selected as the real heat source point.

[0077] Furthermore, in step S11, before converting the surface temperature image from a color image to a grayscale image, the step of equalizing the histogram of the surface temperature image is also included to obtain a surface temperature image with optimized brightness and contrast. The histogram is a statistic of the grayscale distribution of each pixel in the image; histogram equalization is a method for enhancing the contrast of an image, which makes the brightness level of the image smoother by stretching the histogram to make its distribution more uniform.

[0078] Furthermore, in step S11, before converting the surface temperature image from a color image to a grayscale image, the process further includes: performing median filtering on the surface temperature image, specifically, first selecting the size of the filter window, then traversing each pixel in the image in turn, sorting all pixel values ​​in the filter window, and taking the middle value as the new value of the pixel. The new value of the pixel is expressed as:

[0079] P(x,y)=med ian{Q(x+i,y+j)|i,j∈W},

[0080] Among them, P(x,y) represents the new value of the pixel, x and y represent the horizontal and vertical coordinates of the pixel respectively, Q(x+i,y+j) represents the value of the neighborhood around the corresponding pixel in the original image, W represents the filter window, i and j are the parameters that loop within the filter window, and median{} represents the operation of taking the median of the pixel values ​​in the neighborhood.

[0081] In this embodiment, before the surface temperature image is converted from a color image to a grayscale image, the surface temperature image is subjected to histogram equalization and median filtering processing; wherein, the histogram equalization is used for image enhancement. In many cases, due to problems such as lighting and reflection, the dynamic range of the image may be relatively limited, causing the image to be too dim or too bright. In this case, the histogram equalization can expand the dynamic range of the image, enhance the contrast of the image, and make the detail information in the image more obvious. At the same time, the histogram equalization can also make the brightness distribution of the image more uniform. The median filtering process can effectively remove some random, isolated noises, such as salt noise and pepper noise. At the same time, compared with the average filter, the median filter can better protect the edge information of the image while removing noise, avoiding excessive blurring of the image during the denoising process.

[0082] The embodiment of the present application corrects the virtual heat source point on the line connecting the first real heat source point and the second real heat source point by identifying the real heat source point, so that the obtained virtual heat source point can accurately correspond to the first real heat source point and the second real heat source point in the closed isotherm, thereby improving the accuracy of temperature calculation, and only needs to be calculated according to the acquired working temperature data, without understanding the specific physical properties or structure inside the device, so the application cost is low and the scope is wide. It should be noted that since there are multiple points in the closed isotherm that have a great influence on the virtual heat source point, the influence on the test result when the performance of a real heat source point deviates from the standard is relatively smaller, and setting too many thermocouples in a closed isotherm will lead to an increase in real-time cost. Therefore, when the virtual heat source point does not overlap with the real heat source point but the virtual heat source point is not on the line connecting the first real heat source point and the second real heat source point, out of comprehensive consideration between detection accuracy and thermocouple cost, step S14 does not process the virtual heat source point.

[0083] For further information, see Figure 3 The temperature compensation model of each virtual heat source point is fitted according to the ambient surface temperature difference and the test temperature difference, and specifically comprises the steps of:

[0084] Divide the ambient surface temperature difference and the test temperature difference data into a training set and a test set;

[0085] A temperature compensation model is obtained by fitting the training set data using the least squares method; the temperature compensation model is an electrical parameter analysis model;

[0086] The temperature compensation model was tested using the test set data, and the prediction accuracy of the model was evaluated using the mean square error indicator;

[0087] The error curve of the temperature compensation model is drawn, and the overfitting or underfitting phenomenon in the training process of the temperature compensation model is identified through the error curve, and the temperature compensation model is optimized by increasing or decreasing the degree of the polynomial or regularizing the method accordingly.

[0088] The error curve is used to represent the distribution of training error and verification error with the number of iterations. If both the training error and verification error are in a convergence state and the gap between them is small, it means that the temperature compensation model is in a better state; if both the training error and verification error are in a convergence state but the gap between them is large, it is judged that the temperature compensation model is overfitting; if the convergence degree of the training error and verification error is low, it is judged that the temperature compensation model is underfitting.

[0089] This embodiment combines the steps of data set division, least squares modeling, model testing and evaluation, and model optimization to construct and optimize the temperature compensation model of the difference between the equipment environment surface temperature and the test temperature. In the implementation process of this embodiment, first, the approximate relationship between the difference between the environment surface temperature and the test temperature difference is intuitively seen through the scatter plot, and then, by dividing the data into training and test sets, the generalization ability of the model is effectively promoted and overfitting is prevented. Next, the least squares method is used to establish a preliminary model, and the mean square error is used to evaluate the prediction performance of the model. Finally, the overfitting or underfitting phenomenon of the model is identified through the training error curve, so that the temperature compensation model can accurately fit the difference between the environment surface temperature and the test temperature, thereby providing accurate and reliable prediction and evaluation for the temperature compensation of different virtual heat source points, thereby improving the working efficiency and effectiveness of the equipment.

[0090] Another embodiment of the present application provides a surface temperature detection method for a chip-level GaAs power device, based on the above embodiment, further comprising step S6; see Figure 4 , the step S6 specifically comprises the following steps:

[0091] S61. While obtaining the ambient temperature and the operating temperature data of the virtual heat source point in real time in step S5, obtain the thermally sensitive electrical parameters of the power device; taking the IGBT module as an example of the power device, the thermally sensitive electrical parameters at least include threshold voltage, saturation voltage and forward conductance.

[0092] S62. Construct several electrical parameter analysis models, each of which corresponds to a virtual heat source point; wherein the electrical parameter analysis model is a polynomial regression model; the independent variables of the electrical parameter analysis model include thermally sensitive electrical parameters and ambient temperature; and the dependent variable of the electrical parameter analysis model is the operating temperature data of the virtual heat source point.

[0093] S63, training an electrical parameter analysis model through thermal-sensitive electrical parameters, operating temperature data, and thermal-sensitive electrical parameters;

[0094] S64, inputting the real-time acquired thermally sensitive electrical parameters and operating temperature data into the trained electrical parameter analysis model to predict the temperature of the virtual heat source point; when the ratio of the predicted result to the real-time acquired operating temperature data of the virtual heat source point is greater than a preset threshold, it is determined that the power device is faulty.

[0095] This embodiment acquires heat-sensitive electrical parameters and environmental and operating temperature data in real time, constructs an electrical parameter analysis model based on this data, and predicts the temperature of the virtual heat source point through the model. When the deviation between the predicted temperature and the actual temperature exceeds the preset threshold, it can be judged in time that the power device has a fault, thereby improving the monitoring accuracy of the power device and the timeliness of fault detection, thereby achieving the purpose of improving equipment efficiency and extending equipment service life.

[0096] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, the indirect coupling or communication connection of the modules can be electrical, mechanical or other forms.

[0097] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0098] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0099] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-on l ymemory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program codes.

Claims

1. A surface temperature detection method for a chip-level GaAs power device, characterized in that: Includes steps: Acquire a surface temperature image of the device in a working state, and identify several virtual heat source points of the device according to the surface temperature image; A plurality of thermocouples are arranged on the surface of the device, and the thermocouples are used to collect the temperature of the virtual heat source points; the number of the thermocouples is the same as the number of the virtual heat source points; Acquire a temperature image of the device surface at different ambient temperatures, calculate a first test temperature of a virtual heat source point according to the temperature image, and simultaneously acquire a second test temperature of the virtual heat source point through a thermocouple; Calculating a test temperature difference according to the first test temperature and the second test temperature, calculating an ambient surface temperature difference according to the second test temperature and the ambient temperature, and fitting a temperature compensation model for each virtual heat source point according to the ambient surface temperature difference and the test temperature difference; The ambient temperature and the operating temperature data of the virtual heat source point are acquired in real time, and the operating temperature data are corrected according to the ambient temperature and the temperature compensation model.

2. The surface temperature detection method according to claim 1, characterized in that: The method of identifying a plurality of virtual heat source points of a device according to a surface temperature image specifically comprises the steps of: Convert the surface temperature image from a color image to a grayscale image; Set the reference temperature, and draw several closed isotherms in the grayscale image according to the grayscale value corresponding to the reference temperature; The point with the lowest gray value within the closed isotherm is recorded as the virtual heat source point.

3. The surface temperature detection method according to claim 2, characterized in that: The reference temperature is expressed as: T ref =k(T j,max -P×R jc ), Among them, T j,max is the standard maximum junction temperature, R jc is the standard thermal resistance; P is the working power of the device; k is the proportional adjustment coefficient.

4. The surface temperature detection method according to claim 2, characterized in that: The method of identifying a plurality of virtual heat source points of a device according to a surface temperature image further comprises the steps of: The real heat source point in each closed isotherm is identified according to the temperature distribution gradient characteristics; when the virtual heat source point does not overlap with the real heat source point and the virtual heat source point is on the line connecting the first real heat source point and the second real heat source point, the set virtual heat source point is deleted and the first real heat source point and the second real heat source point are respectively set as virtual heat source points; wherein the first real heat source point is the point with the lowest grayscale value among the real heat source points of the closed isotherm; and the second real heat source point is the point with the second lowest grayscale value among the real heat source points of the closed isotherm.

5. The surface temperature detection method according to claim 1, characterized in that: The temperature compensation model for each virtual heat source point is fitted according to the ambient surface temperature difference and the test temperature difference, specifically comprising the steps of: Divide the ambient surface temperature difference and the test temperature difference data into a training set and a test set; A temperature compensation model is obtained by fitting the training set data using the least squares method; the temperature compensation model is an electrical parameter analysis model; The temperature compensation model was tested using the test set data, and the prediction accuracy of the model was evaluated using the mean square error indicator; Draw the error curve of the temperature compensation model, identify the overfitting or underfitting phenomenon in the training process of the temperature compensation model through the error curve, and optimize the temperature compensation model by increasing or decreasing the polynomial degree or regularization method accordingly; the error curve is used to represent the distribution of training error and verification error as the number of iterations changes.

6. The surface temperature detection method according to claim 1, characterized in that: Also includes the steps: Obtain thermally sensitive electrical parameters of power devices; Constructing a plurality of electrical parameter analysis models, each of which corresponds to a virtual heat source point; The electrical parameter analysis model is trained by using the thermal sensitive electrical parameters, the operating temperature data and the thermal sensitive electrical parameters; The real-time acquired thermally sensitive electrical parameters and operating temperature data are input into the trained electrical parameter analysis model to predict the temperature of the virtual heat source point. When the ratio of the predicted result to the real-time acquired operating temperature data of the virtual heat source point is greater than a preset threshold, it is determined that the power device is faulty.

7. The surface temperature detection method according to claim 6, characterized in that: The electrical parameter analysis model is a polynomial regression model; the independent variables of the electrical parameter analysis model include thermally sensitive electrical parameters and ambient temperature; and the dependent variable of the electrical parameter analysis model is the operating temperature data of the virtual heat source point.

8. The surface temperature detection method according to claim 1, characterized in that: The temperature compensation value is expressed as: T i,com =T i,1 -T i,2 , Where i is a positive integer, indicating the sequence number of the virtual heat source point; T i,com is the temperature compensation value of the ith virtual heat source point, T i,1 is the first test temperature of the ith virtual heat source point, T i,2 is the second test temperature of the i-th virtual heat source point; The ambient surface temperature difference is expressed as: T i,def =T i,2 -T s , Among them, T i,def represents the ambient surface temperature difference of the i-th virtual heat source point, T s Indicates the ambient temperature.

9. The surface temperature detection method according to claim 2, characterized in that: Before converting the surface temperature image from a color image to a grayscale image, the method further includes the step of performing equalization processing on the histogram of the surface temperature image to obtain a surface temperature image with optimized brightness and contrast.

10. The surface temperature detection method according to claim 2, characterized in that: Before converting the surface temperature image from a color image to a grayscale image, the step further includes: Select the size of the filter window, traverse each pixel in the image in turn, sort all the pixel values ​​in the filter window, and take the middle value as the new value of the pixel point; the new value of the pixel point is expressed as: P(x,y)=median{Q(x+i,y+j)|i,j∈W}, Among them, P(x,y) represents the new value of the pixel, x and y represent the horizontal and vertical coordinates of the pixel respectively, Q(x+i,y+j) represents the value of the neighborhood around the corresponding pixel in the original image, W represents the filter window, i and j are the parameters that loop within the filter window, and median{} represents the operation of taking the median of the pixel values ​​in the neighborhood.

Images