Ruggedized notebook computer low-temperature starting scene recognition control system and method

Abstract

The invention relates to the technical field of recognition control, and provides a reinforced notebook computer low-temperature starting scene recognition control system and method.Through close cooperation of an environment sensing module, a scene recognition module, a preheating control module and a collaborative optimization module, low-temperature starting challenges can be comprehensively and intelligently dealt with, and the low-temperature starting efficiency is improved. All the modules share data and are matched with one another to form a closed-loop intelligent control system, so that the starting success rate and stability of the reinforced notebook computer in a low-temperature environment are remarkably improved, the service life of the reinforced notebook computer is remarkably prolonged, and the usability and reliability of the reinforced notebook computer in an extreme environment are greatly improved.

Description

Technical Field

[0001] The present application relates to the technical field of identification control, and particularly relates to a recognition control system and method for the low-temperature startup scenario of a rugged notebook computer. Background Art

[0002] In a low-temperature environment, the startup of a rugged notebook computer faces many challenges. Conventional control systems mostly analyze based on a single data source or a simple model, and cannot make full use of historical fault data and complex environmental characteristics. Moreover, the existing technologies usually adopt a unified preheating process, without considering factors such as the resource status and battery characteristics of the rugged notebook computer. At the same time, when planning the preheating path, the problem of hardware thermal stress concentration is not considered, which is likely to cause irreversible damage to the rugged notebook computer. The existing low-temperature startup technologies for rugged notebook computers cannot comprehensively and intelligently cope with complex low-temperature environments. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present application provides a recognition control system and method for the low-temperature startup scenario of a rugged notebook computer.

[0004] In a first aspect, the present application provides a recognition control system for the low-temperature startup scenario of a rugged notebook computer, and the system includes: an environment perception module, a scenario recognition module, a preheating control module, and a collaborative optimization module;

[0005] The environment perception module is used to obtain the perception data of the rugged notebook computer through a non-uniform sampling logic. The perception data includes the physical environment, device status, and user behavior signals, eliminate the time offset of the perception data through time correlation alignment, and construct an environmental feature matrix;

[0006] The scenario recognition module is used to identify the scenario type and confidence score based on the environmental feature matrix and the historical low-temperature fault library through a composite neural network, and generate preheating constraint conditions;

[0007] The preheating control module is used to dynamically plan a preheating instruction set according to the scenario type, confidence score, preheating constraint conditions, and resource status. The preheating instruction set includes different preheating paths, heating temperature sequences, and hardware resource allocations;

[0008] The collaborative optimization module is used to monitor the execution status data generated during the preheating control process according to the different preheating paths, heating temperature sequences, and hardware resources, dynamically balance the heating speed, system stability, and resource efficiency based on the execution status data to optimize the preheating instruction set, and identify the abnormal status of the execution status data to update the historical low-temperature fault library.

[0009] As an optional implementation manner, the non-uniform sampling logic includes:

[0010] Obtain the perception data of the rugged laptop and configure the basic sampling frequency of the perception data;

[0011] Calculate the temperature gradient values of any two regions among the main board, battery and casing, and adjust the basic sampling frequency of the regions according to the temperature gradient values;

[0012] Configure the altitude threshold, compare the altitude obtained by barometric conversion with the altitude threshold, and trigger temperature error correction;

[0013] Identify the user operation intensity according to the operation mode to adjust the basic sampling frequency of the perception data.

[0014] As an optional implementation manner, the logic of time-correlation alignment includes:

[0015] Determine the grouping of the perception data according to the sampling frequency of the perception data, including a high-frequency group and a low-frequency group;

[0016] By assigning different weight factors to the high-frequency group and the low-frequency group, calculate the weighted Euclidean distance of the perception data in the time series in different groups;

[0017] Dynamically adjust the alignment position of the perception data according to the weighted Euclidean distance of the perception data in the time series in different groups.

[0018] As an optional implementation manner, the construction logic of the environmental feature matrix includes:

[0019] Extract the features of the perception data from the perception data that has been time-correlation aligned;

[0020] Select the features of the perception data through a random forest to form a feature subset;

[0021] Weight and fill according to the feature subsets of different time series to form an environmental feature matrix.

[0022] As an optional implementation manner, the recognition logic of the scene type and confidence score includes:

[0023] Expand the dimension of the environmental feature matrix, and integrate the data in the historical low-temperature fault library into the dimension of the environmental feature matrix in an encoded form to form an environmental fault matrix;

[0024] Form a composite neural network through a convolutional neural network and a long short-term memory network, and train the composite neural network with the historical perception data and historical low-temperature fault data of the rugged laptop;

[0025] Input the environmental fault matrix into the trained composite neural network multiple times, and count the scene type with the most occurrences as the recognition result;

[0026] Calculate the difference between the probability of the scene type in the recognition result and the probabilities of other scene types, and combine the weight coefficient determined according to the fluctuation degree of the physical environment to obtain the confidence score.

[0027] As an alternative implementation, the generation logic of the preheating constraint conditions includes:

[0028] Initially set the preheating constraint conditions according to the scene type. The preheating constraint conditions include the maximum allowable heating rate and the battery instantaneous discharge safety threshold;

[0029] Simulate the preheating control process, continuously monitor the real-time device status, and compare it with the initial preheating constraint conditions to obtain a comparison result;

[0030] Aiming at the successful startup of the rugged laptop computer and the minimum battery life loss, adjust the preheating constraint conditions according to the comparison result.

[0031] As an alternative implementation, the planning logic of the preheating instruction set includes:

[0032] Generate an initial preheating instruction set according to the scene type, confidence score, and preheating constraint conditions;

[0033] Combine the remaining battery capacity in the resource status to adjust the heating temperature sequence;

[0034] Combine the temperature differences in each area in the resource status to adjust the differential preheating path;

[0035] Allocate hardware resources in combination with the number of available logic units of the FPGA in the resource status.

[0036] As an alternative implementation, the optimization logic of the preheating instruction set includes:

[0037] Construct a multi-objective optimization model with the heating speed, system stability, and resource efficiency as the target variables;

[0038] Real-time monitor the execution status data and extract the features of the execution status data to obtain the target features;

[0039] Input the target features into the multi-objective optimization model for iterative calculation, so as to optimize the preheating instruction set.

[0040] As an alternative implementation, the update logic of the historical low-temperature fault library includes:

[0041] Calculate the similarity of the data distribution between the real-time monitored execution status data and the set normal status data, and configure a similarity threshold to judge the abnormal status of the execution status data;

[0042] Classify the abnormal status of the execution status data to obtain the abnormal type, so as to record and update the historical low-temperature failure library.

[0043] In a second aspect, the present application provides a method for identifying and controlling the low-temperature startup scenario of a rugged notebook computer. The method includes: obtaining the sensing data of the rugged notebook computer through non-uniform sampling logic, eliminating the time offset of the sensing data through time correlation alignment, and constructing an environmental feature matrix;

[0044] Based on the environmental feature matrix and the historical low-temperature failure library, identify the scenario type and confidence score through a composite neural network, and generate preheating constraint conditions;

[0045] Dynamically plan the preheating instruction set according to the scenario type, confidence score, preheating constraint conditions and resource status. The preheating instruction set includes differential preheating paths, heating temperature sequences and hardware resource allocations;

[0046] Monitor the execution status data generated during the preheating control process according to the differential preheating path, heating temperature sequence and hardware resources, dynamically balance the heating speed, system stability and resource efficiency based on the execution status data to optimize the preheating instruction set, and identify the abnormal status of the execution status data to update the historical low-temperature failure library.

[0047] Compared with the prior art, the beneficial effects of the present application are: through the close cooperation of the environmental perception module, scenario recognition module, preheating control module and collaborative optimization module, it is possible to comprehensively and intelligently cope with the low-temperature startup challenge. Data sharing and mutual cooperation among the modules form a closed-loop intelligent control system, significantly improving the startup success rate, stability of the rugged notebook computer in a low-temperature environment and the service life of the rugged notebook computer, and greatly enhancing the usability and reliability of the rugged notebook computer in extreme environments.

[0048] The environmental perception module can dynamically adjust the sampling frequency according to the temperature gradient, altitude and user operation intensity through non-uniform sampling logic, accurately obtain key information, and improve the pertinence and effectiveness of data acquisition. The time correlation alignment logic effectively eliminates the time offset of the sensing data, ensuring the time consistency of the sensing data; and the constructed environmental feature matrix highlights the key features, providing a high-quality data basis for subsequent scenario recognition, making the control system more sensitive and accurate in perceiving environmental changes.

[0049] The scenario recognition module significantly improves the accuracy of scenario type recognition and the reliability of the confidence score by fusing the environmental feature matrix with the historical low-temperature failure library and performing scenario recognition through a composite neural network. The preheating constraint conditions generated according to different scenario types can provide accurate guidance for preheating control, making the preheating control more in line with the actual needs of the device and reducing the risk of startup failure caused by scenario misjudgment.

[0050] The preheating control module dynamically plans the preheating instruction set according to the scenario type, confidence score, preheating constraint conditions, and resource status, realizing the refined control of the preheating process. Combining resource status such as the remaining battery capacity, temperatures in each area, and the number of available logic units in the FPGA, it flexibly adjusts the heating temperature sequence, differential preheating path, and hardware resource allocation, ensuring both the fast and safe startup of the rugged laptop while taking into account battery life and hardware stability, effectively improving the preheating efficiency and the reliability of the rugged laptop.

[0051] The collaborative optimization module realizes the dynamic balance of heating speed, system stability, and resource efficiency by constructing a multi-objective optimization model, monitoring the execution status data in real time and optimizing it, further improving the performance of the preheating instruction set. At the same time, it can accurately identify abnormal states of the execution status data and update the historical low-temperature fault library, enhancing the self-adaptability and fault prevention ability of the control system, providing a strong guarantee for the long-term stable operation of the rugged laptop. Description of the Drawings

[0052] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0053] Figure 1 is the system flow chart of a rugged laptop low-temperature startup scenario recognition control system provided by an embodiment of the present application;

[0054] Figure 2 is the environmental feature matrix construction logic diagram of a rugged laptop low-temperature startup scenario recognition control system provided by an embodiment of the present application;

[0055] Figure 3 is the recognition logic diagram of the scenario type and confidence score of a rugged laptop low-temperature startup scenario recognition control system provided by an embodiment of the present application;

[0056] Figure 4 is the method flow chart of a rugged laptop low-temperature startup scenario recognition control method provided by an embodiment of the present application. Detailed Embodiments

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present application more obvious and understandable, the following clearly and completely describes the technical solutions in the embodiments of the present application with reference to the drawings in the specification. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments.

[0058] Example 1

[0059] As Figure 1 shown, the embodiment of the present application provides a system flowchart of a control system for identifying the low - temperature startup scenario of a rugged notebook computer. The system includes an environmental perception module, a scenario recognition module, a pre - heating control module, and a collaborative optimization module.

[0060] The environmental perception module is used to obtain the perception data of the rugged notebook computer through non - uniform sampling logic. The perception data includes physical environment, device status, and user behavior signals. The time offset of the perception data is eliminated through time - correlation alignment, and an environmental feature matrix is constructed.

[0061] The physical environment includes temperature, humidity, and air pressure. The device status includes battery impedance. The user behavior signal includes the operation mode. Among them, the temperature includes the main - board temperature, battery temperature, and shell temperature of the rugged notebook computer.

[0062] Specifically, the non - uniform sampling logic includes:

[0063] Obtain the perception data of the rugged notebook computer and configure the basic sampling frequency of the perception data;

[0064] Calculate the temperature gradient values between any two regions of the main - board, battery, and shell, and adjust the basic sampling frequency of the regions according to the temperature gradient values;

[0065] Configure the altitude threshold, compare the altitude obtained by air - pressure conversion with the altitude threshold, and trigger temperature error correction;

[0066] Identify the user operation intensity according to the operation mode to adjust the basic sampling frequency of the perception data.

[0067] In order to obtain the perception data of the rugged notebook computer, multiple sensors are integrated, such as temperature sensors, humidity sensors, air - pressure sensors, battery - impedance sensors, and sensors for monitoring the user operation mode (such as power - button sensors and touch - pad sensors). When the control system is initialized, the basic sampling frequencies are configured for these sensors. For example, the basic sampling frequencies of temperature sensors, humidity sensors, and air - pressure sensors are set to 100Hz, the basic sampling frequency of the battery - impedance sensor is set to 50Hz, and the basic sampling frequency of the user operation - mode sensor is set to 20Hz. This is because under normal circumstances, these frequencies are sufficient to capture the changes in the physical environment, device status, and user behavior signals, and can balance the data acquisition volume and the energy consumption of the control system.

[0068] Every 1 second, the control system reads the temperature data of three areas: the main board, the battery, and the casing, and calculates the temperature gradient value between any two areas. For example, if the temperature of the main board is T1 and the temperature of the battery is T2, then the temperature gradient value between the main board and the battery is ΔT = |T1 - T2|. When ΔT > 5°C, the sampling frequency of the temperature sensor in the area with a larger temperature change is increased. For example, assuming that the temperature gradient value between the main board and the battery exceeds 5°C, the sampling frequencies of the temperature sensors in the main board and battery areas are increased to 1 kHz. This is because there may be abnormal heat transfer or heat dissipation problems in the area with a larger temperature gradient, and increasing the sampling frequency can more accurately monitor the temperature change and detect potential faults in a timely manner.

[0069] For example, the control system pre-sets the altitude threshold to 3000 meters. The pressure sensor continuously collects pressure data and converts the pressure value into altitude through the conversion formula between pressure and altitude. When the calculated altitude is greater than 3000 meters, temperature error correction is triggered. At this time, the control system obtains the attitude information of the rugged laptop through a six-axis gyroscope to determine whether the rugged laptop is in an environment prone to condensation, such as rapid temperature drop and relative humidity > 70%; if it is determined that the rugged laptop is in an environment prone to condensation, the output of the temperature sensor is corrected according to the gyroscope data; this is because in high-altitude areas, the temperature sensor is vulnerable to condensation and generates measurement errors, and this method can improve the accuracy of temperature measurement.

[0070] By analyzing the data of the user operation mode sensor, the user operation intensity is identified. For example, if the user frequently and quickly presses the power button, or the operation frequency of the touchpad increases sharply within a short period of time, it is determined that the user operation intensity is high. At this time, the control system reduces the sampling frequency of non-critical data (such as humidity sensor data, which has little impact on the low-temperature startup of the device during emergency operations) to 10 Hz, or even temporarily freezes the sampling, while increasing the sampling frequency of the sensor directly related to the user operation (such as the power button sensor) to 50 Hz; this is to ensure the priority of data acquisition related to operations and reduce the occupation of system resources by non-critical data during emergency operations by the user, thereby improving the system response speed.

[0071] By adjusting the sampling frequency according to the temperature gradient, altitude, and user operation intensity, key data is collected with emphasis, improving the pertinence and effectiveness of data collection. This enables the subsequently constructed environmental feature matrix to more accurately reflect the actual situation of the device in the low-temperature startup scenario, providing a more reliable data basis for the scenario recognition module. While ensuring the quality of key data acquisition, the sampling frequency of non-critical data is reasonably adjusted to avoid unnecessary data acquisition, reducing the energy consumption of the control system and the data processing burden, and improving the utilization efficiency of the control system resources. This helps the entire control system maintain stable performance during long-term operation and provides guarantee for the continuous operation of subsequent modules.

[0072] Specifically, the logic of time-correlation alignment includes:

[0073] Determine the grouping of perception data according to the sampling frequency of the perception data, including a high-frequency group and a low-frequency group;

[0074] By assigning different weight factors to the high-frequency group and the low-frequency group, calculate the weighted Euclidean distance of the perception data in the time series for different groups;

[0075] Dynamically adjust the alignment position of the perception data according to the weighted Euclidean distance of the perception data in different groups in the time series.

[0076] The control system groups the perception data according to the sampling frequency of each sensor. The sensor data with a sampling frequency greater than 500Hz is classified into the high-frequency group, such as the temperature sensor data after increasing the sampling frequency in the area with a large temperature gradient. The sensor data with a sampling frequency less than or equal to 500Hz is classified into the low-frequency group, such as the data of most sensors with conventional sampling frequencies. Such grouping is to adopt different processing methods for data with different sampling frequencies, improving the efficiency and accuracy of time-correlation alignment.

[0077] Assign different weight factors to the high-frequency group and the low-frequency group respectively. Since the data in the high-frequency group is more sensitive to changes in the device state, a higher weight factor is assigned, such as 0.8. The weight factor of the low-frequency group data is set to 0.2, and the sum of the weight factor of the high-frequency group and the weight factor of the low-frequency group is 1. For each group of data, 10 consecutive data points are selected as a data segment in the time series. For example, for the data segment T = [t1, t2,..t. 10 ,] of the temperature sensor in the high-frequency group and the data segment H = [h1, h2,...h 1,0 of the humidity sensor in the low-frequency group, calculate the weighted Euclidean distance between the two. In this way, the similarity of data with different frequencies in the time series is comprehensively considered. The calculation formula of the weighted Euclidean distance between the two is as follows:

[0078]

[0079] In the formula, d represents the value of the weighted Euclidean distance between the two, i represents the index, and t i represents the data of the i-th temperature sensor, and h i represents the data of the i-th humidity sensor. and respectively represent the means of the data segments of the temperature sensor and the humidity sensor.

[0080] According to the calculated weighted Euclidean distance, dynamically adjust the alignment positions of the sensed data in different groups. If the weighted Euclidean distance between two data segments is small (obtained by comparing with the distance threshold), it indicates that their change trends in the time series are relatively similar, and they can be aligned on the time axis. For example, by moving the timestamps of the data segments, the starting time points of the two data segments are made to coincide; for data segments with a large weighted Euclidean distance, further analyze the reasons for the differences, such as whether there are sensor failures or environmental interferences, etc. This step ensures the temporal consistency of data from different types of sensors and lays a foundation for accurately constructing the environmental feature matrix in the subsequent steps.

[0081] Through time-correlation alignment, the time offsets caused by different sampling frequencies of data from different sensors are effectively eliminated, making all sensed data temporally consistent. This helps to accurately reflect the mutual relationships between different environmental factors and device states at the same moment when constructing the environmental feature matrix, improving the accuracy and reliability of the matrix; the data after time-correlation alignment is more convenient for subsequent feature extraction and analysis. When constructing the environmental feature matrix in the subsequent steps, more valuable feature information can be extracted based on the accurately time-aligned data, providing better-quality data input for the scene recognition module, thereby improving the accuracy and efficiency of scene recognition.

[0082] Specifically, as Figure 2 shown, the construction logic of the environmental feature matrix includes:

[0083] Extract the features of the sensed data from the time-correlation-aligned sensed data;

[0084] Select the features of the sensed data through random forest to form a feature subset;

[0085] According to the feature subsets of different time series, fill them weighted to form the environmental feature matrix.

[0086] Extract various features from the perception data that has been aligned through time correlation. For temperature data, features such as temperature value, temperature gradient, and temperature change rate are extracted; for humidity data, features such as humidity value and humidity change trend are extracted; for battery impedance data, features such as impedance value and impedance phase angle are extracted; for user operation mode data, the operation mode is encoded in vector form, and features such as operation frequency and operation duration are extracted. For example, for temperature data, the temperature change rate is obtained by calculating the temperature difference between two adjacent sampling points in a region divided by the time interval. This step is to mine valuable information for the identification of the device's low-temperature startup scenario from the original perception data.

[0087] Use the random forest algorithm to screen all the extracted features. The random forest algorithm constructs multiple decision trees, conducts multiple random samplings and trainings on the features, and selects features with higher importance scores according to the importance of the features in the decision trees to form a feature subset. For example, after multiple trainings and calculations, features such as temperature gradient, battery impedance phase angle, and user operation frequency are selected as the feature subset. This is because the random forest algorithm can effectively handle high-dimensional data, remove redundant and unimportant features, and improve the efficiency of subsequent data processing and model performance.

[0088] According to the feature subsets of different time series, perform weighted filling to form an environmental feature matrix. The rows of the matrix represent different time sampling points, and the columns represent the selected features. For each feature, different weights are assigned according to its importance for the device's low-temperature startup. For example, the temperature gradient is very important for judging the thermal state of a rugged laptop, and a weight of 0.6 is assigned; the weight of the user operation frequency is set to 0.3; the weight of the humidity change trend is set to 0.1; when filling the environmental feature matrix, multiply the value of each feature by its corresponding weight and fill it into the corresponding position of the environmental feature matrix. Every 10 seconds, update the environmental feature matrix according to the latest perception data. This step ensures that the environmental feature matrix can highlight key features and more accurately reflect the state of the device at different time points.

[0089] Through feature extraction and random forest to select the feature subset, the features that have a key impact on the identification of the rugged laptop's low-temperature startup scenario are highlighted, reducing the data dimension and the complexity of data processing. This enables the environmental feature matrix to more concisely and effectively express the device state and provide more targeted data for the scenario recognition module; while the environmental feature matrix formed by weighted filling represents the data according to the feature importance, improving the information quality of the environmental feature matrix. Based on such an environmental feature matrix for analysis, the scenario recognition module can more accurately identify different low-temperature startup scenario types and confidence scores, providing a reliable basis for generating preheating constraint conditions in the future.

[0090] The scenario recognition module is used to identify the scenario type and confidence score based on the environmental feature matrix and the historical low-temperature fault library, and generate preheating constraint conditions.

[0091] Specifically, as Figure 3 shown, the recognition logic of the scenario type and confidence score includes:

[0092] Expand the dimension of the environmental feature matrix, and incorporate the data in the historical low-temperature fault library into the dimension of the environmental feature matrix in an encoded form to form an environmental fault matrix;

[0093] A composite neural network is composed of a convolutional neural network and a long short-term memory network, and the composite neural network is trained with the historical perception data and historical low-temperature fault data of the rugged laptop;

[0094] Input the environmental fault matrix into the trained composite neural network multiple times, and the scenario type with the most occurrences is used as the recognition result;

[0095] Calculate the difference between the probability of the scenario type in the recognition result and the probabilities of other scenario types, and combine it with the weight coefficient determined according to the fluctuation degree of the physical environment to obtain the confidence score.

[0096] First, expand the dimension of the environmental feature matrix output by the environmental perception module. Assume that the original environmental feature matrix is an m×n matrix, where m is the number of time sampling points and n is the number of features. Extract relevant data from the historical low-temperature fault library, such as the fault time, fault components, and fault types in past similar low-temperature scenarios. Digitally encode this information. For example, represent the fault type using one-hot encoding and normalize the fault time, etc. Then add the encoded data as new columns to the environmental feature matrix to form an m×(n + k) environmental fault matrix, where k is the number of newly added features extracted and encoded from the historical low-temperature fault library. This is to combine historical fault information with the current environmental features, enabling the composite neural network to learn the feature patterns in different scenarios from more abundant data, improving the accuracy of scenario recognition, thus increasing the dimension and information content of the data and providing more comprehensive input data for the subsequent composite neural network.

[0097] Construct a composite neural network composed of a convolutional neural network and a long short-term memory network. The CNN part contains multiple convolutional layers and pooling layers, which are used to extract spatial features in the environmental fault matrix, such as temperature distribution and device hardware layout information. The LSTM part consists of multiple LSTM units and is used to process time series data, such as the temperature change trend in historical perception data and the time series of historical fault occurrences. The composite neural network is trained with the historical perception data and historical low-temperature fault data of the rugged laptop. During the training process, the stochastic gradient descent algorithm is used as the optimizer, and the cross-entropy loss function is selected as the loss function to minimize the difference between the prediction result and the true scenario type. During the training process, the weights and bias parameters of the composite neural network are continuously adjusted to enable the composite neural network to accurately identify different scenario types. This is because CNN is good at processing spatial features, and LSTM is suitable for processing time series data. The combination of the two can make full use of the information in the environmental fault matrix, improve the accuracy of scenario recognition, so that the composite neural network can learn the feature patterns in different scenarios and have the ability to accurately classify the input data.

[0098] Input the environmental fault matrix into the trained composite neural network multiple times (such as 10 times). Each time the network will output a prediction result of a scenario type, and each prediction result represents the possibility of each scenario type in the form of a probability distribution. Count the scenario type that appears most frequently among these 10 prediction results and use it as the final recognition result. This is because through multiple inputs and statistics, the randomness and error of single prediction can be reduced, and the reliability of the recognition result can be improved, so as to obtain a relatively accurate and reliable recognition result of the scenario type, which can reflect the current low-temperature startup scenario of the rugged laptop.

[0099] Calculate the difference between the probability of the scene type in the recognition result and the probabilities of other scene types. For example, assume the recognition result is the "extreme cold standby" scene with a probability of p, and the probabilities of other scene types (such as "high altitude low temperature", "critical startup", "local icing", and "user forced cold startup") are p1, p2, p3, and p4 respectively. Then calculate the difference Δp = p - max(p1, p2, p3, p4). At the same time, determine a weight coefficient according to the fluctuation degree of the physical environment. For example, measure the fluctuation degree by calculating the standard deviation of physical environments such as temperature, humidity, and air pressure. The greater the fluctuation degree, the smaller the weight coefficient. For example, set the weight coefficient to 1 / (1 + standard deviation of the physical environment); then multiply the difference by the weight coefficient to obtain the confidence score. When environmental parameters conflict, such as high humidity causing sensor reading drift, the established dynamic confidence threshold mechanism will automatically detect this situation. At this time, start the redundant sensor cross-validation, that is, re-obtain the perception data through redundant devices such as backup humidity sensors, and input the new perception data into the composite neural network for re-prediction. Adjust the confidence score according to the two prediction results. This is because by considering the probability difference and the fluctuation degree of the physical environment, the credibility of the recognition result can be evaluated more reasonably, and the dynamic confidence threshold mechanism and redundant sensor cross-validation further improve the accuracy and reliability of the confidence score, so as to obtain a confidence score that can reflect the reliability of the recognition result and provide a reference for subsequent decisions.

[0100] Through the above steps, the low-temperature startup scene type of the rugged notebook computer can be accurately identified from the rich perception data, and a reliable confidence score can be obtained, which enables the control system to more accurately understand the current state of the rugged notebook computer and provide more accurate guidance for subsequent preheating control.

[0101] Specifically, the generation logic of the preheating constraint conditions includes:

[0102] Initial set the preheating constraint conditions according to the scene type. The preheating constraint conditions include the maximum allowable heating rate and the battery instantaneous discharge safety threshold;

[0103] Simulate the preheating control process, continuously monitor the real-time device state, and compare it with the initial preheating constraint conditions to obtain the comparison result;

[0104] With the goal of successfully starting the rugged notebook computer and minimizing the battery life loss, adjust the preheating constraint conditions according to the comparison result.

[0105] According to the recognition result of the scenario type, the preheating constraint conditions are initially set. For example, when the scenario type is "extremely cold standby", since the rugged laptop is in a low-temperature static state for a long time, the chemical reaction rate of the battery slows down and the internal resistance increases. To avoid damage to the battery due to over-discharge, the instantaneous discharge safety threshold of the battery is set to 50% of that at normal temperature, and the maximum allowable heating rate is set to increase by 1°C every 10 seconds. For the "critical start" scenario, the rugged laptop needs to quickly enter the working state, so the instantaneous discharge safety threshold of the battery is increased to 70% of that at normal temperature, and the maximum allowable heating rate is set to increase by 3°C every 10 seconds. This is because different scenario types have different requirements for the rugged laptop. Setting the initial preheating constraint conditions according to the scenario type can provide a basic framework for the subsequent preheating control, thus providing preliminary constraint conditions for the preheating control and guiding the subsequent simulation and adjustment processes.

[0106] Through computer simulation technology, the preheating control process of the rugged laptop is simulated. During the simulation process, the real-time device status is continuously monitored, including the actual discharge current of the battery, the temperature changes of each component, and the usage of hardware resources, etc. These real-time monitored data are compared with the initial preheating constraint conditions. For example, it is checked whether the actual discharge current of the battery exceeds the instantaneous discharge safety threshold of the battery, and whether the temperature change rate of each component meets the maximum allowable heating rate, etc., to obtain the comparison result. This is because through simulation, it is possible to evaluate whether the preheating control process is reasonable before actual operation, discover potential problems in a timely manner, and thus be able to find situations that do not conform to the initial constraint conditions in the preheating control process, providing a basis for adjusting the preheating constraint conditions.

[0107] With the goal of successfully starting the rugged laptop and minimizing the battery life loss, the preheating constraint conditions are adjusted according to the comparison result. If it is found during the simulation that the actual discharge current of the battery is close to the instantaneous discharge safety threshold of the battery and the battery temperature rises slowly, it indicates that the current heating rate may not be able to meet the device startup requirements. At this time, the maximum allowable heating rate is appropriately increased, but at the same time, the battery temperature and the temperature changes of other components are closely monitored to ensure that the hardware tolerance limit is not exceeded. If it is found that the temperature change of a certain component is too fast and exceeds the maximum allowable heating rate, the heating power of that component is reduced and the heating strategy is adjusted. This is because on the premise of ensuring the successful startup of the device, it is necessary to minimize the battery life loss and improve the reliability and service life of the rugged laptop, so as to obtain more reasonable preheating constraint conditions that can protect the battery and hardware devices while meeting the startup requirements of the rugged laptop.

[0108] By generating and adjusting warm-up constraint conditions, the warm-up control process can be made more reasonable and optimized, improving the performance and stability of the device during low-temperature startup. Reasonable warm-up constraint conditions provide accurate guidance for the warm-up control module to plan the warm-up instruction set, ensuring that the warm-up instruction set can be formulated according to the actual needs of the device and achieving effective warm-up control.

[0109] The warm-up control module is used to dynamically plan the warm-up instruction set according to the scene type, confidence score, warm-up constraint conditions, and resource status. The warm-up instruction set includes differential warm-up paths, heating temperature sequences, and hardware resource allocations.

[0110] The resource status includes the remaining battery capacity, the temperature of each region, and the number of available logic units in the FPGA;

[0111] Specifically, the planning logic of the warm-up instruction set includes:

[0112] Generate an initial warm-up instruction set according to the scene type, confidence score, and warm-up constraint conditions;

[0113] Adjust the heating temperature sequence in combination with the remaining battery capacity in the resource status;

[0114] Adjust the differential warm-up path in combination with the temperature of each region in the resource status;

[0115] Flexibly allocate hardware resources in combination with the number of available logic units in the FPGA in the resource status.

[0116] Different scene types have different requirements for device startup. In the extremely cold standby scene, the device temperature needs to be slowly and steadily increased to avoid damage to components such as the battery. In the critical startup scene, the startup conditions need to be quickly reached. The confidence score reflects the reliability of scene recognition. When the score is high, the strategy corresponding to the scene type can be more firmly executed. The warm-up constraint conditions limit the heating process from the perspectives of safety and performance, ensuring that the rugged laptop operates within a safe range. Therefore, generating an initial warm-up instruction set by integrating these factors can provide a basic framework for the subsequent warm-up process.

[0117] For example, when the scene type is "extremely cold standby" and the confidence score is relatively high (such as greater than 90%), in combination with a relatively low maximum allowable heating rate and the battery instantaneous discharge safety threshold, the differential warm-up path is initially planned to heat the battery, motherboard, and screen in sequence at a slower speed, and the heating temperature sequence is set to slowly increase the temperature of the battery to -15°C, the motherboard to -10°C, and the screen to -5°C.

[0118] For example, when the scenario type is "critical startup" and the confidence score is relatively high, considering the need to quickly start the device and increase the heating power, a differential preheating path is planned to prioritize ensuring the rapid temperature rise of the battery and the motherboard. The heating temperature sequence is set such that the battery is quickly heated to -10°C, the motherboard is heated to -5°C, and the screen is maintained at 0°C.

[0119] Integrate the differential preheating path, heating temperature sequence, and preliminary hardware resource allocation scheme determined according to different scenario types and preheating constraint conditions. For the hardware resource allocation scheme, for general scenarios, a certain number of FPGA logic units are allocated for heating control to form an initial preheating instruction set. Thus, an initial preheating instruction set that conforms to the current scenario and constraint conditions is generated, providing a clear guiding direction for the subsequent preheating process, avoiding blind heating, and improving the pertinence and efficiency of preheating.

[0120] The remaining battery capacity directly affects the energy that the battery can provide. Overheating in a low battery state will cause the battery to over-discharge, affecting the battery life and even causing the rugged laptop to fail to start properly. In a high battery state, the heating speed can be appropriately increased to improve the startup efficiency. The battery impedance phase angle reflects the electrochemical state inside the battery. By monitoring its changes, the performance of the battery can be understood in real time, thereby adjusting the heating strategy to protect the battery and improve the preheating effect.

[0121] The remaining capacity information of the battery is obtained in real time through the battery management system and fed back to the preheating control module. When the remaining battery capacity is low (for example, below 20%), the target temperature in the heating temperature sequence is reduced to reduce the energy consumption of the battery. For example, the target heating temperature of the battery is reduced from -10°C to -15°C, and at the same time, the heating time is extended to ensure that the battery can safely warm up in a low battery state.

[0122] When the remaining battery capacity is sufficient (for example, above 80%), the target temperature in the heating temperature sequence can be appropriately increased to speed up the preheating speed. For example, the target heating temperature of the battery is increased from -10°C to -5°C.

[0123] Monitor the change of the battery impedance phase angle in real time. When the impedance phase angle increases, it indicates that the chemical reaction rate inside the battery slows down. At this time, appropriately reduce the heating rate to avoid damage to the battery caused by overheating. For example, when the impedance phase angle exceeds 45°, reduce the heating rate by 20%. Adjust the timing and mode of the heating temperature sequence according to the change trend of the battery impedance phase angle. For example, if the impedance phase angle continues to increase, preferentially perform slow heating on the battery for a long time, and then gradually increase the heating speed after the impedance phase angle stabilizes. Thus, dynamically adjust the heating temperature sequence according to the actual state of the battery, which can not only ensure the rapid startup of the device when the battery is fully charged, but also protect the battery when the battery power is low, extend the battery life, and improve the safety and reliability of the preheating process at the same time.

[0124] Temperature differences in each region will cause the generation of thermal stress. Excessive thermal stress will damage the hardware of the rugged notebook computer, affecting the reliability and service life of the rugged notebook computer. By calculating the stress concentration degree under different differential preheating paths and adjusting the path in a timely manner, thermal stress concentration can be avoided and the hardware of the rugged notebook computer can be protected.

[0125] Obtain the temperature information of each region in real time through temperature sensors distributed in different regions of the rugged notebook computer (such as the battery, motherboard, and screen). Establish a thermal-structural coupling model of the rugged notebook computer through finite element analysis software. Input different differential preheating paths (such as battery → motherboard → screen and motherboard → battery → screen, etc.) and the current temperature information of each region into the thermal-structural coupling model, and calculate the stress concentration degree of each region under different differential preheating paths. If the calculated stress concentration degree is greater than 5 MPa, immediately start the path adjustment mechanism. For example, if heating is carried out according to the conventional battery → motherboard → screen path and the stress concentration degree in a certain region of the motherboard exceeds 5 MPa, then re-plan the preheating path as battery → screen → motherboard, or add a temperature buffer stage to this region during the heating process to reduce stress concentration. Thus, effectively avoid the damage caused to the hardware of the rugged notebook computer by thermal stress concentration, improve the reliability and stability of the rugged notebook computer during the preheating process, and extend the service life of the rugged notebook computer.

[0126] FPGA is an important computing resource in the preheating control process. Reasonably allocating its logic units can improve the accuracy and efficiency of heating control. In the case of limited resources, it is necessary to optimize the allocation of resources according to the actual situation, give priority to ensuring the operation of key functions, and at the same time, further reduce power consumption and resource occupancy by dynamically adjusting the disable list, improving the overall performance of the control system.

[0127] Query the number of available logic units in the FPGA in real time through the management interface of the FPGA, and feedback this information to the preheating control module; when the number of available logic units in the FPGA is sufficient, activate the dedicated heating algorithm core of the FPGA, and use its powerful computing power to precisely control the heating process. At the same time, more logic units can be allocated to monitor the temperature and stress changes in each area.

[0128] When the number of available logic units in the FPGA is tight, prioritize the operation of critical functions. For example, skip the GPU self-check process, and use the saved resources for heating control. At the same time, dynamically adjust the allocation of hardware resources according to the confidence score. If the confidence score is high, it means that the scene recognition is reliable, and resources can be allocated more decisively; if the confidence score is low, a conservative resource allocation strategy is adopted, and a certain amount of resources are reserved for possible exception handling; and according to the number of available logic units in the FPGA and the confidence score, dynamically adjust the disable list. For example, when resources are tight and the confidence score is high, temporarily turn off non-critical peripherals such as the fingerprint recognition module and the Bluetooth module to reduce power consumption and resource occupancy; thus realizing the flexible allocation of hardware resources such as FPGA logic units, improving the utilization efficiency of resources, ensuring the smooth progress of the preheating process under limited resources, and at the same time reducing the power consumption of the system.

[0129] The collaborative optimization module is used to monitor the execution status data generated during the preheating control process according to the differential preheating path, heating temperature sequence, and hardware resources, dynamically balance the heating speed, system stability, and resource efficiency based on the execution status data to optimize the preheating instruction set, and identify the abnormal status of the execution status data to update the historical low-temperature fault library.

[0130] Specifically, the optimization logic of the preheating instruction set includes:

[0131] Construct a multi-objective optimization model with the heating speed, system stability, and resource efficiency as the target variables;

[0132] Real-time monitor the execution status data, and extract the features of the execution status data to obtain the target features;

[0133] Input the target features into the multi-objective optimization model for iterative calculation, thereby optimizing the preheating instruction set.

[0134] During the low-temperature startup process of a rugged laptop, the heating speed, system stability, and resource efficiency are interrelated and mutually restrictive factors. Only pursuing the heating speed will sacrifice system stability and resource efficiency, while overemphasizing system stability or resource efficiency will lead to too slow a heating speed. Therefore, it is necessary to construct a multi-objective optimization model to balance these three goals to achieve the optimal overall performance.

[0135] The selection of the genetic algorithm is because this algorithm does not require the objective function to be continuous and differentiable, can handle complex non-linear problems, and is suitable for solving the multi-objective optimization problem in this control system.

[0136] The heating rate, system stability, and resource efficiency are taken as the objective variables. The calculation formula for the heating rate is (target temperature - current temperature) / elapsed time; the calculation formula for system stability is 1 / (temperature fluctuation variance + number of times of hardware stress exceeding the standard); the calculation formula for resource efficiency is (minimum number of logic units required to complete preheating) / (actual number of occupied logic units).

[0137] Combined with the preheating constraint conditions generated by the scenario recognition module, such as the maximum allowable heating rate and the battery instantaneous discharge safety threshold, as the constraint conditions of the multi-objective optimization model; the genetic algorithm is used as the multi-objective optimization algorithm. The genetic algorithm has the advantages of strong global search ability and good parallelism, and can find a better balance among multiple objectives; thus, a multi-objective optimization model that comprehensively considers the heating rate, system stability, and resource efficiency is constructed, providing a theoretical basis and method support for the subsequent optimization of the preheating instruction set; through the global search ability of the genetic algorithm, a better compromise solution can be found among multiple objectives, improving the overall performance of the control system.

[0138] Real-time monitoring of the execution status data is to timely understand the actual situation of the preheating process, so as to adjust the preheating instruction set according to the actual situation; while extracting the target features is to convert a large amount of raw data into key information related to the optimization objectives, reduce the data dimension, and improve the efficiency of the optimization calculation.

[0139] During the preheating control process according to the differential preheating path, heating temperature sequence, and hardware resources, the execution status data is collected in real time through various sensors, including temperature sensors (for monitoring the temperature of each area), current sensors (for monitoring the battery current), and stress sensors (for monitoring the hardware stress), etc.; the collected raw data is preprocessed, including operations such as filtering and noise reduction, to improve the data quality; the target features related to the heating rate, system stability, and resource efficiency are extracted from the preprocessed data, such as calculating the temperature fluctuation variance from the temperature data and counting the number of times of hardware stress exceeding the standard from the hardware stress data; thus, the execution status data of the preheating process can be obtained in real time and accurately, and the key features related to the optimization objectives are extracted from it, providing reliable data support for the subsequent optimization calculation; through data preprocessing and feature extraction, the redundancy and noise of the data are reduced, and the data quality and the efficiency of the optimization calculation are improved.

[0140] Through iterative calculation, the preheating instruction set is continuously optimized to achieve a better balance among the three objectives of heating speed, system stability, and resource efficiency. The iterative process of the genetic algorithm can perform global search in the solution space, avoiding being trapped in local optimal solutions, thereby finding a better preheating instruction set.

[0141] Randomly generate a set of initial solutions as the population of the genetic algorithm. Each solution represents a preheating instruction set, including differential preheating paths, heating temperature sequences, and hardware resource allocation schemes. Input the target features corresponding to each solution into the multi-objective optimization model to calculate the fitness values of each solution for the three objectives of heating speed, system stability, and resource efficiency. Select individuals in the population according to the fitness values, and choose individuals with higher fitness as parents to generate the next generation population. Perform crossover operations on the selected parent individuals to generate new individuals. The crossover operation can exchange part of the genes of the parent individuals to produce new possible solutions. Perform mutation operations on the newly generated individuals to randomly change part of the genes and increase the diversity of the population. Repeat the above fitness evaluation, selection, crossover, and mutation operations until the iteration termination conditions are met, such as reaching the maximum number of iterations or the fitness values converge. Select the individual with the highest fitness from the final population as the optimal solution, and the preheating instruction set corresponding to this solution is the optimized preheating instruction set.

[0142] Through iterative calculation, an optimized preheating instruction set that can achieve a better balance among the three objectives of heating speed, system stability, and resource efficiency is obtained, improving the overall performance of the control system. The global search ability of the genetic algorithm ensures the reliability and effectiveness of the optimization results, avoiding performance losses caused by local optimal solutions.

[0143] Specifically, the update logic of the historical low-temperature fault library includes:

[0144] Calculate the similarity of the data distribution between the real-time monitored execution status data and the set normal status data, and configure a similarity threshold to judge the abnormal status of the execution status data.

[0145] Classify the abnormal status of the execution status data to obtain the abnormal type to record and update the historical low-temperature fault library.

[0146] Timely discovery of the abnormal status of the execution status data helps to take measures in a timely manner to avoid the further expansion of faults and improve the reliability and stability of the control system. By calculating the similarity and setting the threshold, it is possible to objectively and accurately judge whether the execution status data is abnormal, avoiding the subjectivity and uncertainty of manual judgment.

[0147] Calculate the similarity of the data distribution between the execution status data of real-time monitoring and the set normal status data. It can be calculated by the method of Euclidean distance. For example, regard the temperature data of real-time monitoring and the temperature data in the normal state as two vectors and calculate the Euclidean distance between them; configure the similarity threshold. When the calculated similarity is lower than the threshold, it is judged that the execution status data is in an abnormal state. For example, set the similarity threshold to 0.8. When the calculated similarity is 0.7, it is considered that the execution status data is abnormal; thus, it is possible to judge the abnormal state of the execution status data in a timely and accurate manner, providing a basis for subsequent abnormal classification and the update of the historical low-temperature fault library; and through objective similarity calculation and threshold judgment methods, the accuracy and reliability of the abnormal state judgment are improved.

[0148] Classifying the abnormal state helps to better understand the nature and cause of the fault, providing a reference for subsequent fault handling and prevention. Through classification, different types of abnormal states can be sorted out and summarized, facilitating the management and maintenance of the historical low-temperature fault library.

[0149] According to historical experience and system characteristics, define different abnormal types, such as temperature abnormality (including too high or too low), hardware stress abnormality, and battery current abnormality, etc.; according to the characteristics of the execution status data, match the abnormal state with the defined abnormal types. For example, if the temperature in the execution status data exceeds the normal range, it is judged as a temperature abnormality; for some abnormal states that are difficult to accurately classify through feature matching, manual intervention can be introduced and judged and classified by professionals; thus, the abnormal state of the execution status data is accurately classified, providing clear abnormal information for the update of the historical low-temperature fault library, and through classification and sorting, the management efficiency and usability of the historical low-temperature fault library are improved, facilitating subsequent fault analysis and prevention.

[0150] Recording and updating the historical low-temperature fault library can accumulate fault experience during the operation of the control system, providing a reference for subsequent fault prevention and handling. Through regular updates and associations, ensure that the historical low-temperature fault library can timely reflect the latest state of the control system, improving the accuracy and reliability of scenario recognition and preheating control.

[0151] Store the classified abnormal types and corresponding execution status data in the historical low-temperature fault database. A database management system can be used for data storage. Regularly update the historical low-temperature fault database, delete outdated fault data, and add new fault data to ensure the timeliness and accuracy of the historical low-temperature fault database. Then, associate the updated historical low-temperature fault database with the scenario recognition module and the preheating control module, so that the latest fault information can be referenced during subsequent scenario recognition and preheating control processes. As a result, the historical low-temperature fault database is updated in a timely and accurate manner, providing richer and more reliable reference information for fault prevention and handling in the control system. Moreover, through data association, the accuracy and reliability of scenario recognition and preheating control are improved, and the probability of faults occurring is reduced.

[0152] Embodiment 2

[0153] As Figure 4 shown, the present application embodiment provides a method flowchart of a method for identifying and controlling the low-temperature startup scenario of a rugged notebook computer. The method includes:

[0154] Obtain the sensed data of the rugged notebook computer through non-uniform sampling logic, eliminate the time offset of the sensed data through time correlation alignment, and construct an environmental feature matrix.

[0155] Based on the environmental feature matrix and the historical low-temperature fault database, identify the scenario type and confidence score through a composite neural network, and generate preheating constraint conditions.

[0156] Dynamically plan the preheating instruction set according to the scenario type, confidence score, preheating constraint conditions, and resource status. The preheating instruction set includes differential preheating paths, heating temperature sequences, and hardware resource allocations.

[0157] Monitor the execution status data generated during the preheating control process according to the differential preheating paths, heating temperature sequences, and hardware resources, dynamically balance the heating speed, system stability, and resource efficiency based on the execution status data to optimize the preheating instruction set, and identify the abnormal status of the execution status data to update the historical low-temperature fault database.

[0158] Since the principle of the method in the embodiment of the present application for solving problems is similar to that of the above-mentioned system in the embodiment of the present application, the implementation of the method can refer to the implementation of the system, and the repeated parts will not be elaborated.

Claims

1. A recognition control system for the low-temperature startup scenario of a rugged notebook computer, characterized in that, Including: An environmental perception module, a scene recognition module, a preheating control module, and a collaborative optimization module; The environmental perception module is used to obtain the perception data of the rugged laptop through non-uniform sampling logic. The perception data includes physical environment, device status, and user behavior signals. It eliminates the time offset of the perception data through time correlation alignment and constructs an environmental feature matrix; The scene recognition module is used to identify the scene type and confidence score based on the environmental feature matrix and the historical low-temperature fault library through a composite neural network, and generate preheating constraint conditions; The preheating control module is used to dynamically plan a preheating instruction set according to the scene type, confidence score, preheating constraint conditions, and resource status. The preheating instruction set includes a differential preheating path, a heating temperature sequence, and hardware resource allocation; The collaborative optimization module is used to monitor the execution status data generated during the preheating control process according to the differential preheating path, heating temperature sequence, and hardware resources, dynamically balance the heating speed, system stability, and resource efficiency based on the execution status data to optimize the preheating instruction set, and identify the abnormal status of the execution status data to update the historical low-temperature fault library.

2. The recognition control system for low-temperature startup scenarios of a reinforced laptop computer according to claim 1, wherein The non-uniform sampling logic includes: Obtaining the perception data of the rugged laptop and configuring the basic sampling frequency of the perception data; Calculating the temperature gradient values between any two regions of the motherboard, battery, and casing, and adjusting the basic sampling frequency of the regions according to the temperature gradient values; Configuring an altitude threshold, comparing the altitude obtained by barometric conversion with the altitude threshold, and triggering temperature error correction; Identifying the user operation intensity according to the operation mode to adjust the basic sampling frequency of the perception data.

3. The recognition control system for low-temperature startup scenarios of a reinforced laptop computer according to claim 2, wherein The logic of the time correlation alignment includes: Determining the grouping of the perception data according to the sampling frequency of the perception data, including a high-frequency group and a low-frequency group; Calculating the weighted Euclidean distance of the perception data in the time series in different groups by assigning different weight factors to the high-frequency group and the low-frequency group; Dynamically adjusting the alignment position of the perception data according to the weighted Euclidean distance of the perception data in different groups in the time series.

4. The recognition control system for low-temperature startup scenarios of a reinforced laptop computer according to claim 3, wherein, The construction logic of the environmental feature matrix includes: Extracting the features of the perception data from the perception data that has undergone time correlation alignment; Selecting the features of the perception data through a random forest to form a feature subset; Weighted filling according to the feature subsets of different time series to form an environmental feature matrix.

5. The identification control system for low-temperature startup scenarios of a reinforced laptop computer according to claim 4, wherein, The recognition logic of the scene type and confidence score includes: Expanding the dimension of the environmental feature matrix and integrating the data in the historical low-temperature fault library into the dimension of the environmental feature matrix in an encoded form to form an environmental fault matrix; Forming a composite neural network by a convolutional neural network and a long short-term memory network, and training the composite neural network with the historical perception data and historical low-temperature fault data of the rugged laptop; Inputting the environmental fault matrix into the trained composite neural network multiple times, and counting the scene type with the most occurrences as the recognition result; Calculating the difference between the probability of the scene type in the recognition result and the probabilities of other scene types, and combining the weight coefficient determined according to the fluctuation degree of the physical environment to obtain the confidence score.

6. The cold start scenario recognition control system for a reinforced notebook computer according to claim 5, characterized in that, The generation logic of the preheating constraint conditions includes: Initially set preheating constraint conditions according to the scenario type, where the preheating constraint conditions include the maximum allowable heating rate and the battery instantaneous discharge safety threshold; Simulate the preheating control process, continuously monitor the real-time device status, and compare it with the initial preheating constraint conditions to obtain a comparison result; Aim at the successful startup of the rugged laptop and the minimum battery life loss, and adjust the preheating constraint conditions according to the comparison result.

7. The identification control system for low-temperature startup scenarios of a reinforced laptop computer according to claim 6, characterized in that The planning logic of the preheating instruction set includes: Generate an initial preheating instruction set according to the scenario type, confidence score, and preheating constraint conditions; Adjust the heating temperature sequence in combination with the remaining battery capacity in the resource status; Adjust the differential preheating path in combination with the temperature differences in each area of the resource status; Allocate hardware resources in combination with the number of available logic units of the FPGA in the resource status.

8. The identification control system for low-temperature startup scenarios of a reinforced laptop computer according to claim 7, characterized in that, The optimization logic of the preheating instruction set includes: Construct a multi-objective optimization model with the heating speed, system stability, and resource efficiency as the target variables; Real-time monitor the execution status data, and extract the features of the execution status data to obtain the target features; Input the target features into the multi-objective optimization model for iterative calculation to optimize the preheating instruction set.

9. The recognition control system for low-temperature startup scenarios of a reinforced laptop computer according to claim 8, wherein The update logic of the historical low-temperature fault library includes: Calculate the similarity of the data distribution between the real-time monitored execution status data and the set normal status data, and configure a similarity threshold to judge the abnormal status of the execution status data; Classify the abnormal status of the execution status data to obtain the abnormal type, so as to record and update the historical low-temperature fault library.

10. A method for identifying and controlling the low-temperature startup scenario of a rugged notebook computer, which is implemented based on the system for identifying the low-temperature startup scenario of a rugged notebook computer according to any one of claims 1-9, characterized in that, Include: Obtain the perception data of the rugged laptop through the non-uniform sampling logic, eliminate the time offset of the perception data through time correlation alignment, and construct an environmental feature matrix; Based on the environmental feature matrix and the historical low-temperature fault library, identify the scenario type and confidence score through a composite neural network, and generate preheating constraint conditions; Dynamically plan the preheating instruction set according to the scenario type, confidence score, preheating constraint conditions, and resource status. The preheating instruction set includes a differential preheating path, a heating temperature sequence, and hardware resource allocation; Monitor the execution status data generated during the preheating control process according to the differential preheating path, heating temperature sequence, and hardware resources, and dynamically balance the heating speed, system stability, and resource efficiency based on the execution status data to optimize the preheating instruction set, and identify the abnormal status of the execution status data to update the historical low-temperature fault library.

Images