Vehicle-gauge-level chip life prediction method, system and equipment based on deep learning

Abstract

The invention discloses a vehicle gauge level chip life prediction method, system and device based on deep learning, and the method comprises the steps: obtaining first full life cycle data of accelerated aging of a plurality of test chips under a test working condition, and obtaining second full life cycle data of a plurality of sample chips under a real driving working condition; according to the first full life cycle data and a pre-constructed CNN-LSTM attention mechanism, combining network training to obtain a chip life prediction initial model; performing fine tuning training on the chip life prediction initial model according to the second full life cycle data to obtain a chip life prediction target model; temperature time sequence data, electrical time sequence data and environment time sequence data of the target chip under the real driving working condition are obtained and input into the chip life prediction target model, and the remaining service life of the target chip is obtained. The method improves the accuracy and real-time performance of vehicle gauge level chip life prediction, and can be applied to the technical field of artificial intelligence.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, system and device for predicting the lifespan of automotive-grade chips based on deep learning. Background Technology

[0002] Current predictions of automotive-grade chip lifetime primarily rely on physical models to accelerate testing and statistical methods. These methods, based on specific physical failure mechanisms, derive expected lifetimes under normal operating conditions from accelerated experimental data. The Arrhenius equation, the Coffin-Manson model, and the Hallberg-Peck model are three of the most commonly used core models.

[0003] The Arrhenius equation primarily addresses failure mechanisms under high-temperature environments. It uses high-temperature accelerated testing (such as HTOL) to convert the testing time of a chip at accelerated temperature conditions (e.g., 125°C) into equivalent lifespan years at actual application temperatures (e.g., 87°C). This model assumes that chip failure is mainly influenced by temperature-driven chemical reaction processes.

[0004] The Coffin-Manson model focuses on fatigue damage caused by thermomechanical stress, quantifies the impact of temperature cycling on solder joints / packages, and reveals the relationship between the number of cycles and material life. This model is suitable for evaluating high-frequency temperature changes, such as frequent engine start-stop cycles.

[0005] The Hallberg-Peck model simulates the accelerated corrosion effect of high temperature and high humidity environments on solder pads, and is used to evaluate the reliability of chips in humid environments such as automotive electronics.

[0006] Existing methods are effective within a certain range, but they have the following limitations: 1) Models rely on physical assumptions and have poor adaptability: Traditional physical models heavily rely on pre-set failure mechanisms and simplified assumptions (such as single stress dominance, linear damage accumulation, etc.), making it difficult to adapt to the multi-stress coupling effects encountered by automotive-grade chips in actual applications. In actual driving environments, multiple stresses such as temperature, humidity, voltage, current, and vibration coexist and interact, and traditional models cannot accurately capture this complexity.

[0007] 2) Limited data processing capabilities: Existing methods do not adequately utilize real-time operational data and complex operating condition data. Although existing online aging diagnostic methods integrate intelligent sensors, they are mainly based on multi-physics coupled simulation models and lack the ability to mine high-dimensional and nonlinear relationships. Traditional statistical methods struggle to extract valuable lifetime characteristics from massive amounts of real-time monitoring data.

[0008] 3) Insufficient early warning capability: Most existing methods provide average lifespan prediction or macroscopic early warning, which cannot achieve accurate early failure prediction. Automotive-grade chips need to meet the IATF16949 quality management standard of zero failure, but traditional methods have limitations in terms of early warning sensitivity and accuracy, making it difficult to meet the high reliability requirements of automotive electronics.

[0009] 4) High testing costs and long cycles: Traditional accelerated life testing requires a large number of samples and long-term experiments. For example, high-temperature operation life testing usually requires more than 1,000 hours. As the process technology evolves to below 5nm, various effects such as process disturbances, layout effects, and noise are superimposed, causing the testing complexity to increase exponentially and the tape-out cycle to be significantly extended.

[0010] 5) Ignoring individual differences: Most existing methods are based on population statistics and cannot fully account for the individual differences of each chip. Minor deviations in the manufacturing process, material inhomogeneity, and other factors can lead to significant differences in the actual lifespan of chips in the same batch, and traditional methods are difficult to accurately predict for individual chips.

[0011] The above problems urgently need to be addressed. Summary of the Invention

[0012] The purpose of this invention is to at least partially solve one of the technical problems existing in the prior art.

[0013] Therefore, one objective of this invention is to provide a deep learning-based method for predicting the lifespan of automotive-grade chips, which improves the accuracy and real-time performance of lifespan prediction for automotive-grade chips.

[0014] Another objective of this invention is to provide a deep learning-based automotive-grade chip lifetime prediction system.

[0015] To achieve the above-mentioned technical objectives, the technical solutions adopted in the embodiments of the present invention include: On one hand, embodiments of the present invention provide a deep learning-based method for predicting the lifetime of automotive-grade chips, comprising the following steps: Acquire first full lifecycle data of multiple test chips under accelerated aging under test conditions, and acquire second full lifecycle data of multiple sample chips under real driving conditions; An initial model for chip lifespan prediction is obtained by training the first full lifespan data and a pre-built CNN-LSTM attention mechanism joint network. The chip life prediction initial model is fine-tuned and trained based on the second full life cycle data to obtain the chip life prediction target model. Acquire temperature timing data, electrical timing data, and environmental timing data of the target chip under real driving conditions, and input the temperature timing data, electrical timing data, and environmental timing data into the chip lifetime prediction target model to obtain the remaining lifetime of the target chip. The first full life cycle data and the second full life cycle data both include temperature data, electrical data and environmental data.

[0016] Furthermore, in one embodiment of the present invention, the step of training an initial chip lifetime prediction model based on the first full lifecycle data and a pre-built CNN-LSTM attention mechanism joint network specifically includes: Based on the first full lifecycle data, determine the first temperature timing sample, the first electrical timing sample, the first environmental timing sample, and the corresponding first chip lifetime tag; The first temperature time series sample, the first electrical time series sample, and the first environmental time series sample are input into the CNN-LSTM attention mechanism joint network to obtain the first chip lifetime prediction value. A first data fitting loss value is determined based on the first chip lifetime prediction value and the first chip lifetime label. A first physical regularization term is calculated based on the Coffin-Manson model using the first temperature time series sample, the first electrical time series sample, and the first environmental time series sample. Then, a first loss value is determined based on the first data fitting loss value and the first physical regularization term. The parameters of the CNN-LSTM attention mechanism joint network are updated based on the first loss value to obtain the initial model for chip lifetime prediction.

[0017] Further, in one embodiment of the present invention, the CNN-LSTM attention mechanism joint network includes an input layer, a multi-branch CNN layer, a feature fusion layer, an LSTM layer, and an output layer. The step of inputting the first temperature time series sample, the first electrical time series sample, and the first environmental time series sample into the CNN-LSTM attention mechanism joint network to obtain the first chip lifetime prediction value specifically includes: The first temperature time series sample, the first electrical time series sample, and the first environmental time series sample are convolved by the multi-branch CNN layer to obtain temperature feature vector, electrical feature vector, and environmental feature vector, respectively. The feature fusion layer performs feature fusion on the temperature feature vector, the electrical feature vector, and the environmental feature vector based on a multi-head self-attention mechanism to obtain a fused feature vector; The hidden state vector is obtained by calculating the hidden state of the fused feature vector through the LSTM layer; The hidden state vector is mapped to the first chip lifetime prediction value through the output layer.

[0018] Furthermore, in one embodiment of the present invention, the initial chip lifetime prediction model is fine-tuned and trained based on the second full lifecycle data to obtain the target chip lifetime prediction model: Based on the second full lifecycle data, determine the second temperature timing sample, the second electrical timing sample, the second environmental timing sample, and the corresponding second chip lifetime tag; The second temperature time series sample, the second electrical time series sample, and the second environmental time series sample are input into the CNN-LSTM attention mechanism joint network to obtain the second chip lifetime prediction value. The second data fitting loss value is determined based on the second chip lifetime prediction value and the second chip lifetime label. The physical regularization term is calculated based on the Coffin-Manson model based on the second temperature time series sample, the second electrical time series sample and the second environmental time series sample. The second loss value is then determined based on the second data fitting loss value and the second physical regularization term. Based on the first loss value, the parameters of the LSTM layer and output layer of the initial chip lifetime prediction model are fine-tuned to obtain the target chip lifetime prediction model.

[0019] Furthermore, in one embodiment of the present invention, the automotive-grade chip lifetime prediction method further includes: The expected inspection time of the target chip is determined based on the remaining service life, and maintenance decision recommendations for the target chip are generated based on the expected inspection time.

[0020] Furthermore, in one embodiment of the present invention, the automotive-grade chip lifetime prediction method further includes: Obtain the maintenance and inspection results of the target chip, and determine the actual chip life of the target chip based on the maintenance and inspection results; The chip life prediction target model is incrementally trained based on the temperature time series data, the electrical time series data, the environmental time series data, and the actual chip lifespan to obtain the optimized chip lifespan prediction target model.

[0021] On the other hand, embodiments of the present invention provide a deep learning-based automotive-grade chip lifetime prediction system, comprising: The data acquisition module is used to acquire the first full life cycle data of multiple test chips under accelerated aging under test conditions, and to acquire the second full life cycle data of multiple sample chips under real driving conditions. The model training module is used to train an initial model for chip lifespan prediction based on the first full lifespan data and a pre-built CNN-LSTM attention mechanism joint network. The model fine-tuning module is used to fine-tune and train the initial chip life prediction model based on the second full life cycle data to obtain the target chip life prediction model. The prediction module is used to acquire temperature timing data, electrical timing data and environmental timing data of the target chip under real driving conditions, and input the temperature timing data, electrical timing data and environmental timing data into the chip life prediction target model to obtain the remaining life of the target chip. The first full life cycle data and the second full life cycle data both include temperature data, electrical data and environmental data.

[0022] On the other hand, embodiments of the present invention provide an electronic device, the electronic device including a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for implementing communication between the processor and the memory, wherein when the program is executed by the processor, it implements the deep learning-based automotive-grade chip lifetime prediction method described above.

[0023] On the other hand, embodiments of the present invention also provide a vehicle, the vehicle including the deep learning-based automotive-grade chip lifetime prediction system or electronic device as described above.

[0024] On the other hand, embodiments of the present invention also provide a storage medium, which is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, which can be executed by one or more processors to implement the deep learning-based automotive-grade chip lifetime prediction method described above.

[0025] On the other hand, embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the deep learning-based automotive-grade chip lifetime prediction method described above.

[0026] The advantages and beneficial effects of the present invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention: This invention acquires first full-lifecycle data of multiple test chips under accelerated aging conditions and second full-lifecycle data of multiple sample chips under real-world driving conditions. An initial chip lifespan prediction model is trained using the first full-lifecycle data and a pre-built CNN-LSTM attention mechanism joint network. The initial model is then fine-tuned using the second full-lifecycle data to obtain a target model for chip lifespan prediction. Temperature, electrical, and environmental time-series data of the target chip under real-world driving conditions are acquired and input into the target model to obtain the remaining lifespan of the target chip. This invention trains the initial chip lifespan prediction model based on the full-lifecycle data of accelerated aging under test conditions, fine-tunes the initial model based on the full-lifecycle data of the chip under real-world driving conditions to obtain the target model, and then performs real-time lifespan prediction on the target chip based on this target model, improving the accuracy and real-time performance of automotive-grade chip lifespan prediction. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments of the present invention are described below. It should be understood that the drawings described below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A flowchart illustrating the steps of a deep learning-based automotive-grade chip lifetime prediction method provided in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a deep learning-based automotive-grade chip lifetime prediction system provided in an embodiment of the present invention; Figure 3 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0029] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application. It should be noted that although functional modules are divided in the system schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system schematic diagram or the order in the flowchart. The step numbers in the following embodiments are only set for ease of explanation and do not limit the order between steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0030] In the description of this invention, "multiple" means two or more. The use of "first" and "second" is for distinguishing technical features only and should not be construed as indicating or implying relative importance, the number of indicated technical features, or the order of the indicated technical features. Furthermore, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0031] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards of the relevant countries and regions. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirects to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data for the proper functioning of the embodiments of this application obtained.

[0032] like Figure 1 The diagram shows a flowchart of a deep learning-based method for predicting the lifetime of automotive-grade chips according to an embodiment of the present invention. (Refer to...) Figure 1 This invention provides a deep learning-based method for predicting the lifespan of automotive-grade chips, specifically including the following steps: S101. Obtain the first full life cycle data of multiple test chips under accelerated aging under test conditions, and obtain the second full life cycle data of multiple sample chips under real driving conditions. S102. Based on the first full lifecycle data and the pre-built CNN-LSTM attention mechanism joint network, an initial model for chip lifecycle prediction is trained; S103. Fine-tune and train the initial chip life prediction model based on the second full life cycle data to obtain the target chip life prediction model. S104. Obtain the temperature timing data, electrical timing data and environmental timing data of the target chip under real driving conditions, and input the temperature timing data, electrical timing data and environmental timing data into the chip life prediction target model to obtain the remaining life of the target chip. Both the first and second full lifecycle data include temperature data, electrical data, and environmental data.

[0033] Specifically, this invention uses deep learning algorithms to automatically learn the degradation patterns of automotive-grade chips under complex multi-stress environments, breaking through the limitations of simplified assumptions in traditional physical models and achieving accurate modeling of various stress coupling effects such as temperature, voltage, current, and humidity. Utilizing real-time chip operating data, it establishes a dedicated lifespan prediction model for each chip model, overcoming the shortcomings of batch testing in reflecting individual differences and providing customized lifespan assessment and health management for each chip. It designs a multi-dimensional feature extraction and anomaly detection mechanism to achieve early identification of minor chip performance degradation, significantly improving early warning capabilities and meeting the high standard requirement of zero failure for automotive-grade chips. Through the transfer learning capabilities of deep learning models, it reduces reliance on extensive accelerated lifespan testing, significantly reducing verification costs and time, and accelerating the product launch process.

[0034] It can be recognized that the embodiments of the present invention train an initial chip life prediction model based on the full life cycle data of accelerated aging of the chip under test conditions, fine-tunes the initial chip life prediction model based on the full life cycle data of the chip under real driving conditions to obtain a target chip life prediction model, and performs real-time life prediction of the target chip based on the target chip life prediction model, thereby improving the accuracy and real-time performance of automotive-grade chip life prediction.

[0035] Taking the lifespan prediction and management of an automotive power semiconductor device (such as an IGBT module) as an example, this invention provides a detailed explanation of its implementation. Power semiconductor devices are core components of the electric drive system of electric vehicles, and their reliability directly affects the overall vehicle performance and safety. These devices are subjected to electrothermal stress coupling during operation, resulting in a complex degradation mechanism. Traditional methods employ rainflow analysis of temperature cycling or calculation of composite fatigue factors, but these are insufficient for accurately predicting lifespan under actual driving conditions.

[0036] First, multimodal data acquisition was conducted, including temperature data (junction temperature changes), electrical data (operating voltage, current, switching losses, and other parameters), environmental data (ambient temperature and humidity), and historical data (device model, initial test parameters, and previous maintenance records). Full lifecycle data from 200 similar power modules were collected, including normal operation and accelerated aging test data, forming two training datasets.

[0037] As a further optional implementation, an initial chip lifetime prediction model is obtained by training a pre-built CNN-LSTM attention mechanism joint network based on the first full lifecycle data, which specifically includes: S201. Determine the first temperature timing sample, the first electrical timing sample, the first environmental timing sample, and the corresponding first chip lifetime tag based on the first full life cycle data. S202. Input the first temperature time series sample, the first electrical time series sample, and the first environmental time series sample into the CNN-LSTM attention mechanism joint network to obtain the first chip lifetime prediction value. S203. Determine the first data fitting loss value based on the first chip lifetime prediction value and the first chip lifetime label, and calculate the first physical regularization term based on the Coffin-Manson model based on the first temperature time series sample, the first electrical time series sample and the first environmental time series sample, and then determine the first loss value based on the first data fitting loss value and the first physical regularization term. S204. Update the parameters of the CNN-LSTM attention mechanism joint network according to the first loss value to obtain the initial model for chip lifetime prediction.

[0038] Specifically, this embodiment of the invention employs transfer learning technology. First, the model is pre-trained on accelerated aging data, and then fine-tuned on actual driving data to address the scarcity of real-world lifespan data. Simultaneously, this embodiment introduces physical constraints. A data fitting loss term is determined based on the predicted chip lifespan and the chip lifespan label. A physical regularization term is calculated based on the Coffin-Manson model. The training loss value is determined based on the sum of the data fitting loss value and the physical regularization term, thereby updating the model parameters. This embodiment incorporates the Coffin-Manson model as a regularization term into the loss function, guiding the network to learn feature representations that conform to physical laws.

[0039] As a further optional implementation, the CNN-LSTM attention mechanism joint network includes an input layer, a multi-branch CNN layer, a feature fusion layer, an LSTM layer, and an output layer. The first temperature time series sample, the first electrical time series sample, and the first environmental time series sample are input into the CNN-LSTM attention mechanism joint network to obtain the first chip lifetime prediction value, which specifically includes: S301. The first temperature time series sample, the first electrical time series sample, and the first environmental time series sample are convolved by a multi-branch CNN layer to obtain the temperature feature vector, the electrical feature vector, and the environmental feature vector. S302. The feature fusion layer performs feature fusion on the temperature feature vector, electrical feature vector and environmental feature vector based on the multi-head self-attention mechanism to obtain the fused feature vector. S303. The hidden state vector is obtained by calculating the hidden state of the fused feature vector through the LSTM layer. S304. The hidden state vector is mapped to the first chip lifetime prediction value through the output layer.

[0040] Specifically, the multi-branch CNN layer in this embodiment uses three convolutional layers with 32, 64, and 128 filters respectively; the LSTM part uses a two-layer stacked structure with 128 hidden units; and there are eight attention heads. The multi-branch CNN layer performs convolution processing on the first temperature time-series sample, the first electrical time-series sample, and the first environmental time-series sample to obtain temperature feature vectors, electrical feature vectors, and environmental feature vectors respectively; the feature fusion layer fuses the temperature feature vector, electrical feature vector, and environmental feature vector based on a multi-head self-attention mechanism to obtain a fused feature vector; the LSTM layer calculates the hidden state of the fused feature vector to obtain a hidden state vector; and the output layer maps the hidden state vector to a first chip lifetime prediction value.

[0041] As a further optional implementation, the initial chip lifetime prediction model is fine-tuned and trained based on the second full lifecycle data to obtain the target chip lifetime prediction model: S401. Determine the second temperature timing sample, the second electrical timing sample, the second environmental timing sample, and the corresponding second chip lifetime tag based on the second full life cycle data. S402. Input the second temperature time series sample, the second electrical time series sample, and the second environmental time series sample into the CNN-LSTM attention mechanism joint network to obtain the second chip lifetime prediction value. S403. Determine the second data fitting loss value based on the second chip lifetime prediction value and the second chip lifetime label, and calculate the physical regularization term based on the Coffin-Manson model based on the second temperature time series sample, the second electrical time series sample and the second environmental time series sample. Then, determine the second loss value based on the second data fitting loss value and the second physical regularization term. S404. Based on the first loss value, fine-tune the parameters of the LSTM layer and output layer of the initial chip lifetime prediction model to obtain the target chip lifetime prediction model.

[0042] Specifically, when fine-tuning the initial model for chip lifetime prediction, the parameters of the input layer, the multi-branch CNN layer, and the feature fusion layer are fixed, and only the parameters of the LSTM layer and the output layer are updated.

[0043] As an optional implementation, the automotive-grade chip lifetime prediction method further includes: S105. Determine the expected testing time of the target chip based on the remaining service life, and generate maintenance decision recommendations for the target chip based on the expected testing time.

[0044] After the chip lifetime prediction target model is put into actual operation, the monitoring and prediction process for a single power module is as follows: 1) Real-time monitoring: During vehicle operation, the system continuously collects the working data of the power module and generates a data slice for each fixed time period, which includes temperature data, electrical data and environmental data at a certain number of time points.

[0045] 2) Feature extraction: Input the data slices into the trained chip lifetime prediction target model, and extract degradation-related features through the CNN-LSTM attention mechanism of the joint network.

[0046] 3) Lifetime Prediction: Based on the extracted features, predict the remaining lifespan of the power module and provide a confidence interval. Perform corresponding data evaluation, and increase the monitoring frequency when the temperature cycle count approaches the design threshold.

[0047] 4) Maintenance Recommendations: Taking into account the predicted lifespan and vehicle maintenance schedule, the system recommends that the power module be inspected preventively during the next scheduled maintenance (with the expected time given) to avoid sudden failures that could paralyze the entire vehicle.

[0048] As an optional implementation, the automotive-grade chip lifetime prediction method further includes: S106. Obtain the maintenance and testing results of the target chip, and determine the actual chip life of the target chip based on the maintenance and testing results; S107. Based on temperature time series data, electrical time series data, environmental time series data, and actual chip lifespan, incrementally train the chip lifespan prediction target model to obtain the optimized chip lifespan prediction target model.

[0049] Specifically, after performing maintenance and inspection on the target chip, its actual chip lifespan is determined based on the results of the maintenance and inspection. Then, the temperature timing data, electrical timing data, environmental timing data, and actual chip lifespan of the target chip are used as new training samples to incrementally train the chip lifespan prediction target model, thereby achieving continuous optimization of the chip lifespan prediction target model.

[0050] This invention integrates internal chip sensor data, external environmental data, and design attribute data for lifespan prediction, solving the challenges of multi-source heterogeneous data alignment and feature extraction. It embeds a traditional lifespan model as a physical prior into a neural network, designing a physical information-guided deep learning framework to improve model interpretability and extrapolation capabilities. A multi-head self-attention mechanism is optimized for chip lifespan prediction, enabling the model to adaptively focus on critical stress periods and enhance early warning sensitivity. Bayesian deep learning technology is used to provide confidence intervals for prediction results, meeting the requirements of automotive-grade functional safety for uncertainty assessment. A complete closed loop encompassing data acquisition, prediction, verification, and feedback is designed, allowing the model to continuously evolve during use and adapt to individual chip characteristics.

[0051] The method steps of the embodiments of the present invention have been described above. It can be understood that the embodiments of the present invention train an initial chip lifespan prediction model based on the full lifespan data of the chip under accelerated aging conditions during testing. The initial chip lifespan prediction model is then fine-tuned and trained based on the full lifespan data of the chip under real driving conditions to obtain a target chip lifespan prediction model. Based on this target chip lifespan prediction model, the lifespan of the target chip is predicted in real time, thus improving the accuracy and real-time performance of automotive-grade chip lifespan prediction.

[0052] Compared with the prior art, the embodiments of the present invention have the following advantages: 1) Significantly improved prediction accuracy: By fusing multi-source data and deep learning algorithms, this invention can improve lifespan prediction accuracy by more than 30%. Experiments show that, compared with the traditional Arrhenius model, the prediction error of this invention in complex driving environments is reduced to less than 15%, while the error of traditional methods often exceeds 40%.

[0053] 2) Enhanced early warning timeliness: It can identify potential failure risks more than 1,000 hours in advance, making preventative maintenance possible. This is of great value for applications with extremely high safety requirements, such as autonomous driving systems, and can significantly reduce the risk of safety accidents caused by sudden chip failures.

[0054] 3) Personalized Health Management: Achieving personalized prediction based on a "chip-model" approach, fully considering the unique performance degradation trajectory of each chip. This enables the system to formulate optimal maintenance strategies based on the actual condition of the chip, extending chip lifespan by approximately 20% and reducing total lifecycle costs.

[0055] 4) Multi-scenario adaptability: The deep learning model of this invention has strong generalization ability and can adapt to diverse scenarios with different vehicle models, driving styles, and environmental conditions. Whether in hot and humid coastal areas or cold and dry northern regions, the system can provide accurate lifespan predictions.

[0056] 5) Design optimization feedback: By analyzing a large amount of chip lifespan data, this invention can guide chip design optimization in reverse, identify weak links in the reliability of the design stage, provide data support for the next generation of chip design, and form a closed-loop optimization system.

[0057] like Figure 2 The diagram shown is a structural schematic of a deep learning-based automotive-grade chip lifetime prediction system provided in an embodiment of the present invention. (Refer to...) Figure 2 This invention provides a deep learning-based automotive-grade chip lifetime prediction system, comprising: The data acquisition module is used to acquire the first full life cycle data of multiple test chips under accelerated aging under test conditions, and to acquire the second full life cycle data of multiple sample chips under real driving conditions. The model training module is used to train an initial model for chip lifespan prediction based on the first full lifespan data and a pre-built CNN-LSTM attention mechanism joint network. The model fine-tuning module is used to fine-tune and train the initial chip life prediction model based on the second full life cycle data to obtain the target chip life prediction model. The prediction module is used to acquire temperature timing data, electrical timing data and environmental timing data of the target chip under real driving conditions, and input the temperature timing data, electrical timing data and environmental timing data into the chip life prediction target model to obtain the remaining life of the target chip. Both the first and second full lifecycle data include temperature data, electrical data, and environmental data.

[0058] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0059] This invention also provides an electronic device, comprising: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for communication between the processor and the memory. When the program is executed by the processor, it implements the aforementioned deep learning-based automotive-grade chip lifetime prediction method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0060] like Figure 3 The diagram shown is a hardware structure schematic of an electronic device provided in an embodiment of the present invention. (Refer to...) Figure 3 This invention provides an electronic device, comprising: The processor 301 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention. The memory 302 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 302 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 302 and is called and executed by the processor 301 to implement the deep learning-based automotive-grade chip lifetime prediction method of the embodiments of this invention. Input / output interface 303 is used to implement information input and output; The communication interface 304 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 305 transmits information between various components of the device (e.g., processor 301, memory 302, input / output interface 303, and communication interface 304); The processor 301, memory 302, input / output interface 303, and communication interface 304 are connected to each other within the device via bus 305.

[0061] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0062] This invention also provides a vehicle that includes the electric drive assembly of the aforementioned deep learning-based automotive-grade chip lifetime prediction system or electronic device.

[0063] The vehicle can be a private car, such as a sedan, SUV, MPV, or pickup truck. It can also be a commercial vehicle, such as a van, bus, small truck, or large semi-trailer. The vehicle must have an electric motor capable of outputting power or acting as a generator to store mechanical energy. When the vehicle is a new energy vehicle, it can be a hybrid or a pure electric vehicle.

[0064] Since the vehicle applies all the technical solutions of the above-mentioned deep learning-based automotive-grade chip life prediction system or electronic device, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be repeated here.

[0065] This invention also provides a storage medium, which is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, which can be executed by one or more processors to implement the above-mentioned deep learning-based automotive-grade chip lifetime prediction method.

[0066] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0067] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0068] This invention also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform... Figure 1 The method shown.

[0069] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0070] The embodiments described in this invention are for the purpose of more clearly illustrating the technical solutions of the embodiments of this invention, and do not constitute a limitation on the technical solutions provided by the embodiments of this invention. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this invention are also applicable to similar technical problems.

[0071] The terms "first," "second," "third," "fourth," etc. (if present) in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0072] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the aforementioned blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0073] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the aforementioned functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0074] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0075] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0076] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the aforementioned program can be printed, because the aforementioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0077] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0078] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0079] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0080] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A deep learning-based vehicle-grade chip life prediction method, characterized in that, The method comprises the following steps: obtaining first full life cycle data of a plurality of test chips under accelerated aging in a test condition, and obtaining second full life cycle data of a plurality of sample chips under real driving conditions; training an initial chip life prediction model according to the first full life cycle data and a pre-constructed CNN-LSTM attention mechanism combined network; fine-tuning the initial chip life prediction model according to the second full life cycle data to obtain a target chip life prediction model; obtaining temperature time series data, electrical time series data and environmental time series data of a target chip under real driving conditions, inputting the temperature time series data, the electrical time series data and the environmental time series data into the target chip life prediction model to obtain the remaining useful life of the target chip; wherein the first full life cycle data and the second full life cycle data each comprise temperature data, electrical data and environmental data.

2. The deep learning-based automotive-grade chip lifetime prediction method of claim 1, wherein, The initial chip life prediction model is trained according to the first full life cycle data and a pre-constructed CNN-LSTM attention mechanism combined network, which specifically comprises: determining first temperature time series samples, first electrical time series samples, first environmental time series samples and corresponding first chip life labels according to the first full life cycle data; inputting the first temperature time series samples, the first electrical time series samples and the first environmental time series samples into the CNN-LSTM attention mechanism combined network to obtain first chip life prediction values; determining a first data fitting loss value according to the first chip life prediction values and the first chip life labels, and calculating a first physical regularization term based on the Coffin-Manson model according to the first temperature time series samples, the first electrical time series samples and the first environmental time series samples, and then determining a first loss value according to the first data fitting loss value and the first physical regularization term; updating parameters of the CNN-LSTM attention mechanism combined network according to the first loss value to obtain the initial chip life prediction model.

3. The deep learning-based automotive-grade chip lifetime prediction method of claim 2, wherein, The CNN-LSTM attention mechanism combined network comprises an input layer, a multi-branch CNN layer, a feature fusion layer, an LSTM layer and an output layer, and the first chip life prediction values are obtained by inputting the first temperature time series samples, the first electrical time series samples and the first environmental time series samples into the CNN-LSTM attention mechanism combined network, which specifically comprises: performing convolution processing on the first temperature time series samples, the first electrical time series samples and the first environmental time series samples respectively through the multi-branch CNN layer to obtain temperature feature vectors, electrical feature vectors and environmental feature vectors; performing feature fusion on the temperature feature vectors, the electrical feature vectors and the environmental feature vectors based on a multi-head self-attention mechanism through the feature fusion layer to obtain a fusion feature vector; performing hidden state calculation on the fusion feature vector through the LSTM layer to obtain a hidden state vector; map the hidden state vector to the first chip lifetime prediction value through the output layer.

4. The deep learning-based automotive-grade chip lifetime prediction method of claim 3, wherein, The chip lifetime prediction initial model is fine-tuned according to the second full life cycle data to obtain a chip lifetime prediction target model: According to the second full life cycle data, a second temperature time sequence sample, a second electrical time sequence sample, a second environmental time sequence sample and a corresponding second chip lifetime label are determined. The second temperature time sequence sample, the second electrical time sequence sample and the second environmental time sequence sample are input into the CNN-LSTM attention mechanism combined network to obtain a second chip lifetime prediction value. According to the second chip lifetime prediction value and the second chip lifetime label, a second data fitting loss value is determined, and a second physical regularization term is calculated based on the Coffin-Manson model according to the second temperature time sequence sample, the second electrical time sequence sample and the second environmental time sequence sample, and then a second loss value is determined according to the second data fitting loss value and the second physical regularization term. The LSTM layer and the output layer of the chip lifetime prediction initial model are parameter fine-tuned according to the first loss value to obtain the chip lifetime prediction target model.

5. The deep learning-based automotive-grade chip lifetime prediction method according to any one of claims 1 to 4, characterized in that, The vehicle-grade chip lifetime prediction method further comprises: According to the remaining useful life, the expected detection time of the target chip is determined, and a maintenance decision suggestion for the target chip is generated according to the expected detection time.

6. The deep learning-based automotive-grade chip lifetime prediction method of claim 5, wherein, The vehicle-grade chip lifetime prediction method further comprises: Obtain the maintenance detection result of the target chip, and determine the true chip lifetime of the target chip according to the maintenance detection result; According to the temperature time sequence data, the electrical time sequence data, the environmental time sequence data and the true chip lifetime, the chip lifetime prediction target model is incrementally trained to obtain an optimized chip lifetime prediction target model.

7. A deep learning based automotive-grade chip lifetime prediction system, characterized by, It comprises: The data acquisition module is used for acquiring first full life cycle data of a plurality of test chips under accelerated aging test conditions and acquiring second full life cycle data of a plurality of sample chips under real driving conditions; The model training module is used for training a chip lifetime prediction initial model according to the first full life cycle data and a pre-constructed CNN-LSTM attention mechanism combined network; The model fine-tuning module is used for fine-tuning the chip lifetime prediction initial model according to the second full life cycle data to obtain a chip lifetime prediction target model; The prediction module is used for acquiring temperature time sequence data, electrical time sequence data and environmental time sequence data of a target chip under real driving conditions, inputting the temperature time sequence data, the electrical time sequence data and the environmental time sequence data into the chip lifetime prediction target model, and obtaining the remaining useful life of the target chip. The first full life cycle data and the second full life cycle data both include temperature data, electrical data and environmental data.

8. An electronic device, comprising: The electronic device comprises a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing connection communication between the processor and the memory, and the program, when executed by the processor, realizes the deep learning-based vehicle-grade chip life prediction method according to any one of claims 1 to 6.

9. A vehicle characterized by comprising: The vehicle comprises the deep learning-based vehicle-grade chip life prediction system according to claim 7 or the electronic device according to claim 8.

10. A computer program product comprising a computer program, characterized in that, The computer program, when executed by the processor, realizes the deep learning-based vehicle-grade chip life prediction method according to any one of claims 1 to 6.

Images