Chip power consumption detection method and system

Abstract

The invention relates to the technical field of chip power consumption detection, and provides a chip power consumption detection method and system, and the method comprises the steps: obtaining target infrared images of a target chip under different loads through a thermal infrared imager; calculating target power consumption values corresponding to the target chip under different loads; performing first comparison on the target power consumption value and a standard power consumption value, and if the difference value exceeds a first preset range, judging that the power consumption of the target chip is abnormal; if the difference value is within the first preset range, performing second comparison on the target infrared image corresponding to the target power consumption value and the standard infrared image under the current load; and if the result of the second comparison is smaller than the first threshold value, identifying an abnormal area based on the temperature gradient of each pixel point, and marking an abnormal result on the abnormal area according to a preset judgment strategy. The chip power consumption detection process is accelerated, the cost is reduced, the fault identification capability is enhanced, and powerful support is provided for chip quality control.

Description

Technical Field

[0001] The present application relates to the technical field of chip power consumption detection, and in particular, to a chip power consumption detection method and system. Background Art

[0002] The contents of this section merely provide background information related to the present application and may not constitute prior art.

[0003] Chip power consumption detection is a systematic process that uses high-precision electronic measurement equipment to monitor and record the current and voltage values ​​of the chip in real time under various operating modes, and then calculates based on these data to accurately evaluate the energy consumption of the chip under different operating conditions. This process is crucial for verifying the energy efficiency of chip design, identifying potential power consumption issues, and optimizing overall performance, aiming to ensure that the chip achieves low power consumption and high efficiency while meeting performance requirements.

[0004] In the prior art, a Chinese patent with the authorization announcement number CN115078965A discloses a temperature field distribution detection method and device, medium, and chip for a chip. The core of the method is to first determine the operating temperature target values ​​of multiple key nodes on the chip. These key nodes represent the key areas where heat accumulation or temperature changes may occur in the chip under a specific working state. Subsequently, by controlling a series of power consumption simulation units, power consumption simulation is performed one by one according to the operating temperature target values ​​corresponding to these key nodes. This process is intended to simulate the heat generation of the chip in actual work. In order to achieve accurate temperature measurement, the temperature measurement unit is used to sense the temperature conducted by the power consumption simulation unit. There is a specific positional relationship between these temperature measurement units and the power consumption simulation unit. Based on the sensed temperature data and positional relationship, the position-temperature relationship curve of each power consumption simulation unit can be further generated. This step provides key data for the subsequent temperature field distribution analysis. Finally, based on these position-temperature relationship curves, the temperature field distribution of the chip to be tested area can be constructed, thereby realizing the dynamic positioning of the temperature field based on the working characteristics of the chip.

[0005] However, although this method can theoretically provide accurate detection of the overall temperature field distribution of the chip, and uses built-in temperature measurement circuits and power consumption simulation circuits as hardware implementations to reduce dependence on external detection equipment, it has significant defects in actual operation. Specifically, this method requires setting a large number of key nodes for the chip, and equipping each node with a corresponding power consumption simulation unit for simulation one by one. This process not only greatly increases the complexity of the detection process, but also leads to a surge in the number of required equipment. Under the combined effect of these factors, the detection efficiency is reduced, making this method face greater challenges in practical applications. Therefore, how to simplify the detection process, reduce the number of required equipment, and improve detection efficiency while ensuring detection accuracy has become a problem that needs to be solved urgently. Summary of the invention

[0006] In order to solve the above technical problems, the purpose of this application is to provide a chip power consumption detection method and system, which effectively improves the accuracy and efficiency of chip power consumption detection through infrared imaging technology and neural network prediction, and reduces the use of detection equipment.

[0007] The purpose of this application is achieved through the following technical solutions: In a first aspect, the present invention provides a method for detecting chip power consumption, comprising: Obtain target infrared images of the target chip under different loads through an infrared thermal imager; The target infrared images under different loads are input into the calculation model to obtain the target power consumption values ​​corresponding to the target chip under different loads; the calculation model is built based on a neural network model, and is trained by a training set and a validation set containing infrared images of the target chip under various load conditions and their corresponding actual power consumption data; the training set and the validation set are screened from a data set, and the data set uses an infrared thermal imager to perform infrared imaging of the target chip under different load conditions to obtain multiple infrared images, and uses a power consumption detector to measure and record the actual power consumption data corresponding to each infrared image; A first comparison is made between the target power consumption value and the standard power consumption value. If the difference exceeds a first preset range, the target chip power consumption is determined to be abnormal. If the difference is within the first preset range, a second comparison is made between the target infrared image corresponding to the target power consumption value and the standard infrared image under the current load. If the result of the second comparison is that the similarity is greater than or equal to a first threshold, the target chip power consumption is determined to be qualified. If the result of the second comparison is less than the first threshold, an abnormal area where the temperature change exceeds a second preset range is identified based on the temperature gradient of each pixel point of the target infrared image, and an abnormal result is marked on the abnormal area according to a preset judgment strategy. The judgment strategy includes: if the difference between the temperature of the abnormal area and the temperature of the surrounding area is positive, and the difference is greater than a second threshold, the abnormal area is marked as a short circuit; if the difference between the temperature of the abnormal area and the temperature of the surrounding area is negative, and the absolute value of the difference is greater than the second threshold, the abnormal area is marked as an open circuit.

[0008] Furthermore, the step of obtaining target infrared images of the target chip under different loads by using an infrared thermal imager specifically includes: Calibrate the infrared thermal imager; Preheating the target chip under different load conditions; Under each load condition, multiple infrared images are acquired and averaged to obtain the target infrared image to reduce random errors.

[0009] Furthermore, the step of identifying an abnormal area where the temperature change exceeds a second preset range based on the temperature gradient of each pixel point of the target infrared image specifically includes: The target infrared image is identified by using an abnormal point recognition model to obtain the coordinate data of the target abnormal point; the abnormal point recognition model includes: using the Sobel operator to calculate the temperature gradient of the target infrared image in the x direction and the y direction respectively, and calculating the gradient amplitude and gradient direction of each pixel point according to the temperature gradient. If the gradient amplitude of any pixel point exceeds the preset third threshold, it is marked as a possible abnormal point; detecting the gradient direction difference between the possible abnormal point and the adjacent pixel point, and if the gradient direction difference exceeds the preset angle threshold, it is marked as a target abnormal point; Check other pixel points in the preset neighborhood around each target outlier point. If there are other outliers in the preset neighborhood, merge the target outlier point and the other outliers in the preset neighborhood into the same outlier area.

[0010] Furthermore, the expression of the outlier recognition model is:

[0011] in, and Represent the points on the target image exist and The gradient in direction, is the preset third threshold, is the preset angle threshold, for Neighboring pixels, and Represent the points on the target image exist and Directional gradient.

[0012] Furthermore, after marking the abnormal result on the abnormal area according to the preset judgment strategy, it also includes: Obtain a topological map of the target chip as a reference map; According to the location information on the topological map, the abnormal area and its corresponding abnormal result are mapped to the topological map; the location of the abnormal area is marked in a visual manner on the topological map, and the abnormal result is displayed.

[0013] Furthermore, the judgment strategy also includes: If the difference between the temperature of the abnormal area and the temperature of the surrounding area is less than the second threshold, and the temperature of the abnormal area fluctuates periodically, it is marked as an overheated or poorly heat-dissipating area.

[0014] Furthermore, the second comparison adopts one of a structural similarity index comparison or a histogram comparison.

[0015] In a second aspect, the present invention provides a chip power consumption detection system, comprising: A target infrared image acquisition module is used to acquire target infrared images of the target chip under different loads through an infrared thermal imager; The target power consumption value calculation module is used to input the target infrared images under different loads into the calculation model to obtain the target power consumption values ​​corresponding to the target chip under different loads; the calculation model is built based on the neural network model, and is trained by a training set and a verification set containing infrared images of the target chip under various load conditions and their corresponding actual power consumption data; the training set and the verification set are screened from the data set, and the data set uses an infrared thermal imager to perform infrared imaging of the target chip under different load conditions to obtain multiple infrared images, and uses a power consumption detector to measure and record the actual power consumption data corresponding to each infrared image; The abnormality identification module is used to perform a first comparison between the target power consumption value and the standard power consumption value. If the difference exceeds a first preset range, the target chip power consumption is determined to be abnormal; if the difference is within the first preset range, the target infrared image corresponding to the target power consumption value is compared with the standard infrared image under the current load for a second time; if the result of the second comparison is that the similarity is greater than or equal to a first threshold, the target chip power consumption is determined to be qualified; if the result of the second comparison is less than the first threshold, the abnormal area where the temperature change exceeds the second preset range is identified based on the temperature gradient of each pixel point of the target infrared image, and the abnormal result is marked on the abnormal area according to a preset judgment strategy; The judgment strategy includes: if the difference between the temperature of the abnormal area and the temperature of the surrounding area is positive, and the difference is greater than a second threshold, the abnormal area is marked as a short circuit; if the difference between the temperature of the abnormal area and the temperature of the surrounding area is negative, and the absolute value of the difference is greater than the second threshold, the abnormal area is marked as an open circuit.

[0016] In a third aspect, the present invention provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps corresponding to the method in the first aspect are implemented.

[0017] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps corresponding to the method in the first aspect.

[0018] In summary, the technical solution of the embodiment of the present application has at least the following advantages and beneficial effects: The present invention uses an infrared thermal imager to capture infrared images of a target chip under different load conditions. These images reflect the heat distribution of the chip, and thus indirectly reflect the power consumption state of the chip. Subsequently, these infrared images are input into a neural network model that has been pre-trained by a training set and a verification set that contain infrared images of the target chip under various load conditions and their measured power consumption data. The model can calculate the power consumption value of the target chip under different loads. This process not only realizes the rapid detection of power consumption, but also significantly reduces the dependence on professional power consumption detection instruments, thereby reducing the detection cost. By comparing the calculated target power consumption value with the standard power consumption value, and performing similarity analysis on the target infrared image and the standard infrared image, it is possible to accurately determine whether the chip power consumption is abnormal. If the power consumption is abnormal and there is an area with abnormal temperature change in the infrared image, the possible short circuit or open circuit fault of the chip is identified and marked according to the temperature gradient of the abnormal area and the temperature difference of the surrounding area, thereby improving the detection efficiency and accuracy. The chip power consumption detection process is accelerated, the cost is reduced, and the ability of fault identification is enhanced, providing strong support for chip quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of a chip power consumption detection method provided by the present invention; Figure 2 A schematic diagram of the structure of a chip power consumption detection system provided by the present invention; Figure 3 The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0021] A method for detecting chip power consumption proposed in an embodiment of the present application includes: S101, using an infrared thermal imager to obtain infrared images of the target chip under different loads; specifically, using an infrared thermal imager to obtain infrared images of its surface temperature distribution. The first step in this process is to strictly calibrate the infrared thermal imager, which is a key step to ensure that the infrared images obtained in subsequent analysis can accurately reflect the actual temperature distribution of the target chip. The calibration work usually includes adjusting parameters such as the sensitivity, resolution, and temperature range of the thermal imager to match the temperature measurement requirements of the target chip, thereby eliminating the influence of instrument errors on the accuracy of temperature measurement.

[0022] Next, the target chip is preheated under different load conditions. The purpose is to make the temperature distribution of the chip reach a relatively stable state and reduce the measurement error caused by temperature fluctuations. The specific time and method of preheating treatment should be determined according to the chip characteristics and load conditions to ensure that the chip can reach a relatively constant temperature field before formal measurement.

[0023] Under each preset load condition, infrared images are collected. In order to improve the reliability of the data and reduce random errors, multiple shots are taken for each load state, and these images are averaged. Averaging can smooth out random noise in the image and improve the accuracy of temperature measurement, thereby obtaining the target infrared image under the load condition. Ensuring the high quality and representativeness of the image data used for subsequent analysis lays a solid foundation for the accurate calculation of power consumption values.

[0024] S102, inputting the target infrared images under different loads into the calculation model to obtain the target power consumption values ​​corresponding to the target chip under different loads; the calculation model is constructed based on a neural network model, and is trained by a training set and a validation set containing infrared images of the target chip under various load conditions and their corresponding actual power consumption data; the training set and the validation set are screened from a data set, and the data set obtains multiple infrared images by performing infrared imaging of the target chip under different load conditions with an infrared thermal imager, and uses a power consumption detector to measure and record the actual power consumption data corresponding to each infrared image.

[0025] Specifically, infrared images are used as key input data and fed into a pre-built computational model. This computational model is designed based on a neural network architecture. Its advantage is that it can handle complex nonlinear relationships and has the ability to extract key features from input images to predict target power consumption values. The training process of the model is crucial. It relies on a large data set obtained through designed experiments: infrared imaging of the target chip under various load conditions is performed using an infrared thermal imager to capture multiple infrared images reflecting the temperature distribution on the chip surface. At the same time, the actual power consumption data corresponding to each infrared image is accurately measured and recorded using a power consumption detector, so that each pair of infrared images and power consumption data constitutes a training sample.

[0026] To ensure the accuracy and generalization ability of the model, the data set is further divided into a training set and a validation set. The training set is used for the initial learning and parameter adjustment of the model, while the validation set is used to evaluate the performance of the model and prevent overfitting. Through multiple iterations of training, the weights and biases of the neural network are continuously adjusted until the model performs optimally on the validation set and can accurately map the infrared image to the corresponding power consumption value.

[0027] The successful implementation of this step depends not only on high-quality input data, but also on the rationality of model design and the rigor of the training process. Once the model training is completed, it can efficiently predict the power consumption value of the target chip under different load conditions from the new infrared image, providing strong support for subsequent power consumption evaluation and anomaly detection.

[0028] In practical applications, a model based on convolutional neural network (CNN) is used to predict power consumption. CNN has excellent performance in the field of image processing and can automatically extract key features in images. In order to train this CNN model, a dataset containing infrared images of target chips under various load conditions and their corresponding actual power consumption data was prepared. Specifically, we used 1,000 infrared images under different loads as the training set, which covered the temperature distribution of the chip under various working conditions. At the same time, the actual power consumption data corresponding to each image was measured and recorded using a power consumption detector. In order to verify the performance of the model, 200 infrared images were prepared as a validation set.

[0029] During the training process, the following key parameters were set: The network architecture uses a CNN model with multiple convolutional layers and fully connected layers. The convolutional layers are used to extract local features in the image, while the fully connected layers are used to combine these features for power consumption prediction. The learning rate is set to 0.001 to ensure the learning speed of the model while avoiding overfitting. The batch size is set to 32, which means that the model processes 32 images at the same time in each iteration. A sufficient number of iterations (for example, 100) is set to ensure that the model can fully learn the data in the training set.

[0030] During the training process, the cross entropy loss function is used to measure the difference between the model prediction value and the actual power consumption value, and the weights and biases of the model are updated through the back propagation algorithm. As the training progresses, the performance of the model on the validation set gradually stabilizes and eventually reaches the required accuracy.

[0031] S103, the target power consumption value is compared with the standard power consumption value for the first time, if the difference exceeds the first preset range, the target chip power consumption is determined to be abnormal; if the difference is within the first preset range, the target infrared image corresponding to the target power consumption value is compared with the standard infrared image under the current load for the second time; if the result of the second comparison is that the similarity is greater than or equal to the first threshold, the target chip power consumption is determined to be qualified; if the result of the second comparison is less than the first threshold, the abnormal area whose temperature change exceeds the second preset range is identified based on the temperature gradient of each pixel point of the target infrared image, and the abnormal result is marked on the abnormal area according to the preset judgment strategy; the judgment strategy includes: if the difference between the temperature of the abnormal area and the temperature of the surrounding area is positive, and the difference is greater than the second threshold, the abnormal area is marked as a short circuit; if the difference between the temperature of the abnormal area and the temperature of the surrounding area is negative, and the absolute value of the difference is greater than the second threshold, the abnormal area is marked as a break. In addition, if the difference between the temperature of the abnormal area and the temperature of the surrounding area is less than the second threshold, it means that there is no problem with the power consumption of the target chip itself, and the temperature of the abnormal area has periodic fluctuations, which means that there is a problem with the heat dissipation of the position, so it is marked as an overheated or poor heat dissipation area.

[0032] Among them, the first comparison is intended to quickly screen out chips with obviously abnormal power consumption values. If the difference between the target power consumption value and the standard power consumption value exceeds the first preset range set in advance, it can be directly determined that the power consumption of the target chip is abnormal, and there is no need to perform detailed analysis in subsequent steps. However, if the difference between the target power consumption value and the standard power consumption value is within the first preset range, that is, the power consumption value is preliminarily judged to be within the normal range, it is necessary to further check the power consumption state of the chip through image comparison. At this time, the target infrared image corresponding to the target power consumption value is compared with the standard infrared image under the current load conditions for the second comparison, and the second comparison adopts one of the structural similarity index comparison or histogram comparison. The power consumption state of the chip is evaluated by comparing the similarity between the two. The basis for judging the similarity is the preset first threshold. If the similarity between the target infrared image and the standard infrared image is greater than or equal to the first threshold, it means that the power consumption distribution of the chip is consistent with the standard state, and the power consumption of the target chip is determined to be qualified.

[0033] If the result of the second comparison shows that the similarity is less than the first threshold, it indicates that there is an abnormality in the power consumption distribution of the chip, and further analysis of the abnormal area is required. At this time, the temperature gradient of each pixel in the target infrared image is used to identify abnormal areas where the temperature change exceeds the second preset range. For these abnormal areas, detailed analysis and marking are performed according to the preset judgment strategy. The judgment strategy mainly includes two situations: one is that when the temperature of the abnormal area is higher than that of the surrounding area, and the temperature difference is greater than the second threshold, it is determined that the abnormal area may have a short circuit problem; the other is that when the temperature of the abnormal area is lower than that of the surrounding area, and the absolute value of the temperature difference is greater than the second threshold, it is determined that the abnormal area may have a short circuit problem. Through such a judgment strategy, the abnormal power consumption area of ​​the chip can be accurately located, and the corresponding abnormal result mark can be given, which provides strong support for subsequent troubleshooting and repair.

[0034] Among them, the step of identifying an abnormal area whose temperature change exceeds a second preset range based on the temperature gradient of each pixel point of the target infrared image includes: using an abnormal point recognition model to recognize the target infrared image and obtain coordinate data of the target abnormal point; the abnormal recognition model includes: using a Sobel operator to calculate the temperature gradient of the target infrared image in the x direction and the y direction respectively, and calculating the gradient amplitude and gradient direction of each pixel point according to the temperature gradient, if the gradient amplitude of any pixel point exceeds a preset third threshold value, it is marked as a possible abnormal point; detecting the gradient direction difference between the possible abnormal point and the adjacent pixel point, if the gradient direction difference exceeds a preset angle threshold, it is marked as a target abnormal point.

[0035] Check other pixel points in the preset neighborhood around each target outlier point. If there are other outliers in the preset neighborhood, merge the target outlier point and the other outliers in the preset neighborhood into the same outlier area.

[0036] Specifically, the target infrared image is first processed using the Sobel operator to calculate the temperature gradient of the image in the x-direction and y-direction. The Sobel operator is a discrete differential operator that can effectively extract edge information in an image and is used here to capture the gradient of temperature change. By calculating the gradient amplitude and gradient direction of each pixel, the intensity and directionality of the temperature change can be quantified. The specific calculation process is as follows:

[0037] in, for The temperature value at for The temperature value at for The temperature value at for The temperature value at and Points Temperature gradient in the x-direction and the y-direction.

[0038] Gradient Amplitude for:

[0039] Gradient direction for

[0040] If the gradient amplitude of a pixel exceeds the preset third threshold , the point is initially marked as a possible abnormal point, which means that the temperature change at this point is relatively drastic, and may hide power consumption abnormalities. That is:

[0041] Subsequently, in order to further improve the accuracy of recognition, the possible abnormal points marked initially are further verified. By detecting the gradient direction difference between the possible abnormal point and the adjacent pixel point, if the difference exceeds the preset angle threshold, the point is confirmed to be a true abnormal point, that is, the target abnormal point. This step is intended to eliminate those misjudged points caused by noise or error, and ensure that the identified abnormal points have significant temperature gradient changes, and this change has a certain degree of consistency in direction. The specific calculation formula is as follows:

[0042] in, is the gradient direction difference between the possible abnormal point and the adjacent pixel point, is the gradient direction of adjacent pixels, are adjacent pixels, is the angle threshold.

[0043] Combining formulas (1) to (7), the expression of the outlier recognition model is:

[0044] in, and Represent the points on the target image exist and The gradient in direction, is the preset third threshold, is the preset angle threshold, for Neighboring pixels, and Represent the points on the target image exist and Directional gradient.

[0045] After the target anomaly points are determined, the region merging process is required. Check other pixels in the preset neighborhood around each target anomaly point. If there are other marked anomalies in the neighborhood, merge these anomalies into a unified anomaly region. This helps to identify the specific location of the power consumption anomaly and integrate those adjacent but independent anomaly points to form a more complete and accurate description of the anomaly region.

[0046] Based on the same inventive concept, the present invention provides a chip power consumption detection system, comprising: The target infrared image acquisition module 201 is used to acquire target infrared images of the target chip under different loads through an infrared thermal imager.

[0047] The target power consumption value calculation module 202 is used to input the target infrared images under different loads into the calculation model to obtain the target power consumption values ​​corresponding to the target chip under different loads; the calculation model is constructed based on a neural network model, and is trained by a training set and a verification set containing infrared images of the target chip under various load conditions and their corresponding actual power consumption data; the training set and the verification set are obtained by screening the data set, and the data set uses an infrared thermal imager to perform infrared imaging of the target chip under different load conditions to obtain multiple infrared images, and uses a power consumption detector to measure and record the actual power consumption data corresponding to each infrared image.

[0048] The abnormality identification module 203 is used to perform a first comparison between the target power consumption value and the standard power consumption value. If the difference exceeds the first preset range, the target chip power consumption is determined to be abnormal; if the difference is within the first preset range, the target infrared image corresponding to the target power consumption value is compared with the standard infrared image under the current load for a second time; if the result of the second comparison is that the similarity is greater than or equal to the first threshold, the target chip power consumption is determined to be qualified; if the result of the second comparison is less than the first threshold, the abnormal area where the temperature change exceeds the second preset range is identified based on the temperature gradient of each pixel point of the target infrared image, and the abnormal result is marked on the abnormal area according to the preset judgment strategy; The judgment strategy includes: if the difference between the temperature of the abnormal area and the temperature of the surrounding area is positive, and the difference is greater than a second threshold, the abnormal area is marked as a short circuit; if the difference between the temperature of the abnormal area and the temperature of the surrounding area is negative, and the absolute value of the difference is greater than the second threshold, the abnormal area is marked as an open circuit.

[0049] Based on the same inventive concept, the present invention provides an electronic device, including: a memory 302, a processor 301, and a computer program stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program, a method for detecting chip power consumption is implemented.

[0050] Based on the same inventive concept, the present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a method for detecting chip power consumption is implemented.

[0051] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting chip power consumption, characterized in that: include: Obtain target infrared images of the target chip under different loads through an infrared thermal imager; The target infrared images under different loads are input into a calculation model to obtain the target power consumption values ​​corresponding to the target chip under different loads; the calculation model is constructed based on a neural network model, and is trained by a training set and a validation set containing infrared images of the target chip under various load conditions and their corresponding actual power consumption data; the training set and the validation set are obtained by screening a data set, and the data set is obtained by performing infrared imaging of the target chip under different load conditions by an infrared thermal imager to obtain multiple infrared images, and using a power consumption detector to measure and record the actual power consumption data corresponding to each infrared image; Performing a first comparison between the target power consumption value and the standard power consumption value, and if the difference exceeds a first preset range, determining that the power consumption of the target chip is abnormal; If the difference is within the first preset range, performing a second comparison between the target infrared image corresponding to the target power consumption value and the standard infrared image under the current load; If the result of the second comparison is that the similarity is greater than or equal to the first threshold, the power consumption of the target chip is determined to be qualified; if the result of the second comparison is less than the first threshold, an abnormal area where the temperature change exceeds a second preset range is identified based on the temperature gradient of each pixel point of the target infrared image, and an abnormal result is marked on the abnormal area according to a preset judgment strategy; The judgment strategy includes: if the difference between the temperature of the abnormal area and the temperature of the surrounding area is positive, and the difference is greater than a second threshold, marking the abnormal area as a short circuit; If the difference between the temperature of the abnormal area and the temperature of the surrounding area is negative, and the absolute value of the difference is greater than a second threshold, the abnormal area is marked as a short circuit.

2. A chip power consumption detection method according to claim 1, characterized in that: The step of obtaining the target infrared image of the target chip under different loads by using the infrared thermal imager specifically includes: Calibrate the infrared thermal imager; Preheating the target chip under different load conditions; Under each load condition, multiple infrared images are acquired and averaged to obtain the target infrared image to reduce random errors.

3. The method for detecting chip power consumption according to claim 1, characterized in that: The step of identifying an abnormal area where the temperature change exceeds a second preset range based on the temperature gradient of each pixel point of the target infrared image specifically includes: The target infrared image is identified by using an abnormal point recognition model to obtain coordinate data of the target abnormal point; the abnormal point recognition model includes: using the Sobel operator to calculate the temperature gradient of the target infrared image in the x direction and the y direction respectively, and calculating the gradient amplitude and gradient direction of each pixel point according to the temperature gradient, and if the gradient amplitude of any pixel point exceeds a preset third threshold, it is marked as a possible abnormal point; detecting the gradient direction difference between the possible abnormal point and the adjacent pixel point, and if the gradient direction difference exceeds a preset angle threshold, it is marked as a target abnormal point; Check other pixel points in a preset neighborhood around each target abnormal point. If there are other abnormal points in the preset neighborhood, merge the target abnormal point and other abnormal points in the preset neighborhood into the same abnormal area.

4. A chip power consumption detection method according to claim 3, characterized in that: The expression of the outlier recognition model is: middle, and Represent the points on the target image exist and The gradient in direction, is the preset third threshold, is the preset angle threshold, for Neighboring pixels, and Represent the points on the target image exist and Directional gradient.

5. The method for detecting chip power consumption according to claim 1, characterized in that: After marking the abnormal result on the abnormal area according to the preset judgment strategy, the method further includes: Acquire a topological map of the target chip as a reference map; According to the position information on the topological map, the abnormal area and its corresponding abnormal result are mapped to the topological map; the position of the abnormal area is marked in a visual manner on the topological map, and the abnormal result is displayed.

6. The method for detecting chip power consumption according to claim 1, characterized in that: The judgment strategy also includes: If the difference between the temperature of the abnormal area and the temperature of the surrounding area is less than a second threshold, and the temperature of the abnormal area fluctuates periodically, it is marked as an overheated or poorly heat-dissipating area.

7. The method for detecting chip power consumption according to claim 1, characterized in that: The second comparison adopts one of structural similarity index comparison and histogram comparison.

8. A chip power consumption detection system, characterized in that: include: A target infrared image acquisition module is used to acquire target infrared images of the target chip under different loads through an infrared thermal imager; A target power consumption value calculation module is used to input the target infrared images under different loads into a calculation model to obtain the target power consumption values ​​corresponding to the target chip under different loads; the calculation model is constructed based on a neural network model, and is trained by a training set and a verification set containing infrared images of the target chip under various load conditions and their corresponding actual power consumption data; the training set and the verification set are obtained by screening a data set, and the data set is obtained by performing infrared imaging of the target chip under different load conditions by an infrared thermal imager to obtain multiple infrared images, and using a power consumption detector to measure and record the actual power consumption data corresponding to each infrared image; an abnormality identification module, used for first comparing the target power consumption value with the standard power consumption value, and if the difference exceeds a first preset range, determining that the power consumption of the target chip is abnormal; If the difference is within the first preset range, performing a second comparison between the target infrared image corresponding to the target power consumption value and the standard infrared image under the current load; If the result of the second comparison is that the similarity is greater than or equal to the first threshold, the power consumption of the target chip is determined to be qualified; if the result of the second comparison is less than the first threshold, an abnormal area where the temperature change exceeds a second preset range is identified based on the temperature gradient of each pixel point of the target infrared image, and an abnormal result is marked on the abnormal area according to a preset judgment strategy; The judgment strategy includes: if the difference between the temperature of the abnormal area and the temperature of the surrounding area is positive, and the difference is greater than a second threshold, marking the abnormal area as a short circuit; If the difference between the temperature of the abnormal area and the temperature of the surrounding area is negative, and the absolute value of the difference is greater than a second threshold, the abnormal area is marked as a short circuit.

9. An electronic device, characterized in that: The electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, steps corresponding to the method for detecting chip power consumption according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps corresponding to the method for detecting chip power consumption as described in any one of claims 1 to 7 are implemented.

Images