Performance test system of EMMC particle controller

Abstract

The invention discloses a performance test system of an EMMC particle controller, and relates to the technical field of embedded system performance test and prediction, and the system comprises a preprocessing module which collects the operation data of the EMMC particle controller and the historical operation record data of a user, and carries out the preprocessing; the feature extraction module is used for performing feature extraction on the operation data and the user historical operation record data to obtain an operation feature vector and a user feature vector; the performance prediction module is used for constructing an EMMC performance prediction model to obtain a fault mode and a performance prediction value of the EMMC; the user analysis module recognizes a user behavior mode according to the user feature vector, and the operation scene module recognizes a key operation scene based on the user behavior mode and the fault mode. According to the method, the EMMC performance prediction model is constructed, fault prediction based on multi-dimensional features is realized, and especially for nonlinear behavior prediction under the condition of high load or abnormal operation, the false alarm rate and the missing report rate are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of embedded system performance testing and prediction, and in particular to a performance testing system for an EMMC particle controller. Background Art

[0002] EMMC is the abbreviation of Embedded Multi Media Card, also known as an embedded multimedia controller, which belongs to a packaged storage solution that integrates a storage chip, a controller, and an interface. As an efficient storage solution, the embedded multimedia controller is widely used in smart phones, tablet computers, and other mobile devices. With the popularization of mobile devices and the increasing complexity of functions, the performance requirements for EMMC particle controllers are also increasing day by day.

[0003] Although certain progress has been made in existing EMMC performance testing, there are still deficiencies. First of all, there is a lack of in-depth analysis of user behavior patterns. Most traditional testing methods are based on predefined operation scenarios and do not fully consider the personalized needs and behavior habits of users. This results in a large difference between the test results and the actual user experience and cannot comprehensively reflect the performance of the EMMC in the real usage environment. In addition, secondly, the fault prediction ability of existing models is limited. In most cases, the fault prediction model only relies on static running characteristics for modeling, ignoring the changing trends of running characteristics over time and their complex relationships with each other. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a performance testing system for an EMMC particle controller to solve the problem of lack of in-depth analysis of user behavior patterns.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: The present invention provides a performance testing system for an EMMC particle controller, which includes: Collect the running data of the EMMC particle controller and the historical operation record data of the user, and perform preprocessing; Extract features from the running data and the historical operation record data of the user to obtain a running feature vector and a user feature vector; Build an EMMC performance prediction model to obtain the fault mode and performance prediction value of the EMMC; Identify the user behavior pattern according to the user feature vector; Identify the key operation scenario based on the user behavior pattern and the fault mode; Develop a personalized test plan based on key operation scenarios; Execute the personalized test plan and monitor the execution process; Generate a test report according to the execution results.

[0007] As a preferred solution of the performance test system for the EMMC particle controller described in the present invention, wherein: the operation data includes temperature data, voltage data, and read / write speed data; The user's historical operation record data includes operation frequency, operation type, operation time, and operation result.

[0008] As a preferred solution of the performance test system for the EMMC particle controller described in the present invention, wherein: the preprocessing specifically includes the following steps Perform noise point removal, missing value filling, and data normalization processing on all the collected data.

[0009] As a preferred solution of the performance test system for the EMMC particle controller described in the present invention, wherein: feature extraction is performed on the operation data and the user's historical operation record data to obtain an operation feature vector and a user feature vector, specifically including the following steps Use the fast Fourier transform to extract the high-dimensional features of the operation data to form an operation feature vector; Use the sliding window method to capture the number of operations of the operation frequency; Statistical proportion of different operation types; Calculate the time interval between each operation; Statistical proportion of successful and failed operations; Combine the number of operations, the operation proportion, the time interval of the operation, and the proportion of successful and failed operations to form a user feature vector.

[0010] As a preferred solution of the performance test system for the EMMC particle controller described in the present invention, wherein: construct an EMMC performance prediction model to obtain the performance prediction value of the EMMC, specifically including the following steps Define the fault mode according to the operation feature vector; Perform a non-linear mapping on the operation feature vector to obtain an operation feature transformation vector; Perform a linear combination of the operation feature vector, the weight matrix, and the operation feature transformation vector to obtain the calculation result of the linear prediction, and apply the Sigmoid function to normalize the calculation result of the linear prediction; Perform a dot product calculation on the operation feature vector and the non-linear weight vector, and introduce a non-linear transformation to obtain a non-linear adjustment term; By combining the prediction result of the linear part and the non-linear adjustment term, obtain the fault mode score of the EMMC, and its expression is: ; Among them, represents the score of the th fault mode, represents the Sigmoid function, represents the operating feature vector, represents the transpose of the operating feature vector, represents the weight matrix of the th fault mode, represents the operating feature transformation vector, represents the weight of the non - linear adjustment term for the th fault mode, represents the non - linear weight vector; The Softmax function is used to convert the fault mode score into a fault probability;

[0011] Based on the fault probability, the performance prediction value of the EMMC is obtained.

[0012] As a preferred solution of the performance test system for the EMMC particle controller described in the present invention, wherein: obtaining the performance prediction value of the EMMC specifically includes the following steps, Calculate the weighted average of the operating feature vector; Using an optimization algorithm combined with the weighted average of the operating feature vector, calculate the fault probability to obtain the performance prediction value of the EMMC, and its expression is: ; Among them, represents the performance prediction value of the EMMC particle controller at time , represents the number of features participating in the calculation, represents the th normalized operating feature, represents the weighted average of the normalized operating features, represents the square of the standard deviation of the normalized operating features, represents the total number of fault modes, represents the th weight of the fault mode, represents the regularization parameter, represents the th fault probability of the fault mode at time .

[0013] As a preferred solution of the performance test system for the EMMC particle controller described in the present invention, wherein: identifying the user behavior pattern and preference analysis according to the user feature vector specifically includes the following steps, Using the clustering cost function, select the initial cluster centers from the user feature vectors and calculate the Euclidean distance from each data point to the cluster centers; Based on the distances from each data point to the cluster centers, assign the data points to the nearest clusters; After completing the cluster assignment, calculate the center of each cluster as the mean of all data points belonging to that cluster until the clustering cost function reaches its minimum value, and its expression is: ; Where, represents the clustering cost function at the -th iteration, represents the total number of clusters, represents the -th iteration, -th data point set of the -th cluster, represents a data point, represents the feature vector of the -th data point, represents the center of the -th cluster at the -th iteration;

[0014] The cluster centers are the feature representations of the user behavior patterns, and the cluster memberships are the behavior patterns to which the user behavior features belong.

[0015] As a preferred solution of the performance testing system for the EMMC particle controller according to the present invention, wherein: identify the key operation scenarios based on the user behavior patterns and the failure probabilities, and specifically include the following steps, Set the minimum support and the minimum confidence; Convert the operation records of each user into a transaction format; The transaction format contains all operations within a period of time; For each user behavior pattern, apply the FP-Growth algorithm to mine the frequent item sets and the association rules; Use statistical analysis to find the operation scenarios that affect the performance of the EMMC particle controller from the mined association rules; Define a weight function based on the behavior patterns and the failure probabilities, and calculate the priority of each operation scenario; According to the calculated priorities, sort all the key operation scenarios and select H as the high-priority key operation scenarios.

[0016] As a preferred solution of the performance testing system for the EMMC particle controller according to the present invention, wherein: formulate a personalized test plan based on the key operation scenarios, and specifically include the following steps, Build a test framework and configure a fault injection tool; Design specific test cases according to high-priority critical operation scenarios and gradually expand to other test scenarios.

[0017] As a preferred solution of the performance test system for the EMMC particle controller described in the present invention, wherein: generating the test report specifically includes the following steps, Summarize all test logs and performance monitoring data into a central database; Use data analysis tools to analyze the test data and generate charts and statistical reports.

[0018] The beneficial effects of the present invention are as follows: Using a clustering cost function for user behavior pattern recognition, accurately capturing users' personalized needs and behavior habits, providing an important basis for formulating personalized test plans, improving the consistency between test results and actual user experiences. The personalized test plan formulated in this way ensures the flexibility and adaptability of the test plan, can dynamically adjust test strategies according to the needs and operation scenarios of different users, and improves the overall performance and user experience. In addition, an EMMC performance prediction model is constructed to achieve fault prediction based on multi-dimensional features, significantly improving the accuracy of fault prediction, especially for non-linear behavior prediction under high-load or abnormal operating conditions, and reducing the false alarm rate and missed alarm rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a schematic diagram of the performance test system for the EMMC particle controller in Embodiment 1.

[0021] Figure 2 It is a schematic diagram of personalized plan generation in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings in the specification.

[0023] Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0024] Secondly, the "one embodiment" or "embodiment" mentioned herein refers to specific features, structures or characteristics that may be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not all refer to the same embodiment, nor is it an individual or selectively mutually exclusive embodiment with other embodiments.

[0025] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a performance test system for an EMMC particle controller, including the following steps: A preprocessing module that collects the operation data of the EMMC particle controller and the user's historical operation record data, and performs preprocessing; Collect the temperature data of the EMMC particle controller to monitor whether the EMMC overheats during operation and ensure that it operates within a safe temperature range.

[0026] Collect the voltage data of the EMMC particle controller to evaluate the stability of the power supply and prevent performance degradation or hardware damage caused by voltage fluctuations.

[0027] Collect the read and write speed data of the EMMC particle controller to measure the response time and throughput of the EMMC under different operating conditions and evaluate its performance.

[0028] Furthermore, it provides accurate data support to help analyze the performance of the EMMC under different load conditions and provide a basis for optimizing the design.

[0029] Count the frequency of users performing specific operations to understand the users' common operation habits.

[0030] Record the types of operations performed by users, such as file transfer, application startup, etc., and identify the key operations with high frequencies.

[0031] Record the timestamp of each operation to analyze the users' behavior patterns and operation time periods.

[0032] Record the success or failure of each operation to evaluate the reliability and success rate of the operation.

[0033] Furthermore, through detailed user operation records, the usage habits and preferences of different users can be identified, providing data support for formulating personalized test plans.

[0034] Unify and store all data in a central database for subsequent data processing and analysis.

[0035] Use the Z-score method to identify and eliminate abnormal data points to ensure the authenticity and reliability of the data.

[0036] The interpolation method is used to fill in the missing values in the data to ensure the integrity of the data.

[0037] Min-Max normalization is used to convert data of different scales to the same scale, eliminate the dimensional difference, and facilitate subsequent feature extraction and model construction.

[0038] Furthermore, through detailed data collection and preprocessing steps, not only the comprehensive monitoring and capture of the operating status of the EMMC particle controller and the user behavior pattern are realized, but also the consistency, integrity, and reliability of the data are ensured through centralized management and preprocessing.

[0039] The feature extraction module extracts features from the operating data and the user historical operation record data to obtain the operating feature vector and the user feature vector; The fast Fourier transform is used to convert the time series operating data (such as temperature, voltage, read / write speed, etc.) to the frequency domain to capture the periodic and non-periodic components therein. Identify the high-frequency and low-frequency features hidden in the time series data, and these features correspond to different types of fault modes or performance bottlenecks.

[0040] Furthermore, the frequency domain features can capture the complex patterns that are difficult to intuitively discover in the time series data, and improve the model's ability to identify different types of faults.

[0041] The sliding window method can be used to dynamically monitor the user's operation frequency in different time periods and capture the short-term and long-term change trends.

[0042] The user's historical operations are divided into different types (such as file transfer, application startup, read / write operations, etc.), and the proportion of each operation is counted. According to the operation ratio, evaluate which operations have the greatest impact on the performance of the EMMC particle controller and determine the key points of the test.

[0043] By calculating the time interval between each operation, analyze the time distribution of the user's operations and identify the user's behavior patterns.

[0044] Statistical analysis of the success rate of each operation to understand which operations are most likely to fail and identify potential problem points.

[0045] Combine the features of multiple dimensions such as the number of operations, operation ratio, time interval, and success rate to form a comprehensive user feature representation.

[0046] Furthermore, through the fusion of multi-dimensional features, a detailed user portrait is formed, providing a scientific basis for personalized testing.

[0047] The performance prediction module constructs an EMMC performance prediction model to obtain the fault mode and performance prediction value of the EMMC. The specific steps are as follows: Define the fault modes according to the operation feature vectors, classify the possible fault types, and define the corresponding feature representations for them to ensure that all possible fault modes are considered, which improves the generalization ability of the model.

[0048] Through non-linear mapping, capture the complex non-linear relationships between operation features and combine them into an operation feature transformation vector.

[0049] Furthermore, non-linear mapping can better capture dynamic changes and complex behaviors.

[0050] Set up a weight matrix based on the operation feature vectors and operation feature transformation vectors to measure the importance of different operation features for specific fault modes.

[0051] According to the operation feature vectors, operation feature transformation vectors and weight matrix, obtain the calculation result of linear prediction through linear combination, and use the Sigmoid function to normalize the linear prediction result to the range of [0,1] for subsequent probability conversion.

[0052] Set up an initial non-linear weight vector and define the loss function; Minimize the loss function through the convergence of the optimization algorithm to obtain the non-linear weight vector; Quantify the relationship between the operation feature vector and the non-linear weight vector through dot product calculation to obtain the non-linear adjustment term; Furthermore, the non-linear adjustment term helps to capture subtle changes and differences, improving the model's understanding ability of complex scenarios.

[0053] By combining the prediction result of the linear part and the non-linear adjustment term, obtain the fault mode score of the EMMC, and its expression is: ; Among them, represents the score of the th fault mode, represents the Sigmoid function, represents the operation feature vector, represents the transpose of the operation feature vector, represents the th weight matrix of the fault mode, represents the operation feature transformation vector, represents the weight of the non-linear adjustment term for the th fault mode, represents the non-linear weight vector.

[0054] The score of the fault mode The value range of depends on When Greater than zero, its value range is [0, +∞); when equals zero, its value range is [0, 1], when is less than zero, its value range is (-∞, 1]. According to the requirements of specific business scenarios, adjust the magnitude.

[0055] Use the Softmax function to convert the fault mode score into a fault probability, and its expression is: ; where, represents the probability of the th fault mode under the condition of the given operation feature vector , and N represents the total number of fault modes.

[0056] For the convenience of calculation and representation, convert to .

[0057] Normalize each operation feature and calculate the weighted average value of the operation features. Its expression is: ; where, is the weighted average value of the normalized operation features, represents the number of features, represents the weight of the th normalized operation feature, represents the th normalized operation feature.

[0058] Use the optimization algorithm combined with the weighted average value of the operation feature vector to calculate the fault probability, and obtain the performance prediction value of the EMMC. Its expression is: ; where, represents the performance prediction value of the EMMC particle controller at time , represents the number of features participating in the calculation, represents the th normalized operation feature, represents the weighted average value of the normalized operation features, represents the square of the standard deviation of the normalized operation features, represents the total number of fault modes, represents the weight of the th fault mode, represents the regularization parameter, represents the th fault mode at time fault probability.

[0059] This performance prediction value reflects the overall health status and potential risks of the EMMC particle controller at time .

[0060] Set a health threshold, and obtain the health status of the EMMC particle controller based on the range where the performance prediction value is located within the health threshold. The setting of the health threshold is customized according to the usage scenarios and ranges.

[0061] User analysis module, which identifies user behavior patterns according to user feature vectors.

[0062] Randomly select the initial cluster centers from the user feature vectors; For the Euclidean distance from each data point to all the initial cluster centers, and assign each data point to the nearest cluster based on the Euclidean distance.

[0063] Further explanation, by minimizing the distance from each data point to the nearest cluster center, ensure that the data points are reasonably assigned to the most similar clusters, improving the accuracy of clustering.

[0064] For each cluster, calculate its new center as the mean of all the data points belonging to that cluster.

[0065] Further explanation, by continuously updating the cluster centers, gradually optimize the clustering results, making the cluster centers more accurately reflect the characteristics of the data points within the clusters.

[0066] Define a clustering cost function to measure the quality of the current clustering results, and its expression is: ; where represents the clustering cost function at the th iteration, represents the total number of clusters, represents the th iteration, the th cluster's data point set, represents the data point, represents the th data point's feature vector, represents the th iteration, the th cluster's center.

[0067] When the clustering cost function reaches the minimum value, it is considered that the cluster centers and cluster assignments have converged.

[0068] The cluster center represents the characteristic representation of the user behavior pattern, while the cluster membership reflects the behavior pattern to which the user behavior characteristics belong.

[0069] Furthermore, by minimizing the clustering cost function, it is ensured that the final clustering result reaches the optimal state, improving stability and reliability.

[0070] The operation scenario module identifies key operation scenarios based on the user behavior pattern and the failure mode.

[0071] Set the minimum support and support threshold for mining frequent itemsets and association rules to ensure that the mined rules have statistical significance.

[0072] Specifically, by setting the minimum support, those item sets or rules with extremely low frequencies in the dataset can be excluded, ensuring that the mined patterns have statistical significance and practical application value.

[0073] The minimum support threshold is set to filter out unimportant or uncommon patterns and control the computational complexity at the same time. A lower support threshold usually mines more frequent itemsets, but it may contain many meaningless patterns, generate more rules, and the quality of the rules is lower.

[0074] Therefore, by comparing the effects of the thresholds, a balance point is found, which can not only ensure that enough meaningful patterns are mined, but also control the computational complexity. Moreover, an appropriate threshold can significantly reduce the number of candidate item sets, improve the execution efficiency of the algorithm, and reduce the computational cost.

[0075] Convert the operation records of each user into a transaction format, and each transaction contains all operations within a period of time. Transaction formatting helps to better organize and process operation data and facilitate subsequent association rule mining.

[0076] Based on each user behavior pattern, apply the FP-Growth algorithm to mine frequent itemsets and association rules. FP-Growth is an efficient frequent pattern mining algorithm that can quickly find frequent itemsets in large-scale datasets.

[0077] Use statistical analysis methods to find operation scenarios that affect the performance of the EMMC particle controller from the mined association rules. For example, certain operation combinations may lead to a decrease in EMMC performance or an increase in the failure occurrence frequency.

[0078] Calculate the occurrence frequency weight based on the number of data points within the user behavior pattern, input the characteristic representation of each user behavior pattern into the performance prediction model, and calculate its corresponding failure probability; Multiply the occurrence frequency of each behavior pattern by the corresponding failure probability to obtain the influence weight of all behavior patterns;

[0079] Sort the influence weights of the obtained behavior patterns and output the feature representation of the behavior patterns as the key operation scenarios.

[0080] For example, high-frequency write operations and long-duration high-load random reads.

[0081] Select H from the key operation scenarios as the high-priority key operation scenarios. The number of high-priority key operation scenarios is customized according to the actual situation.

[0082] Test scenario module, formulate personalized test scenarios based on the key operation scenarios.

[0083] Establish a test environment, including a hardware platform, software environment, monitoring tools, and logging; Establish a test framework to support batch execution of test cases and improve test efficiency. Ensure that the test framework has a high degree of configurability, allowing users to adjust test parameters (such as load intensity, operation frequency, etc.) as needed.

[0084] Select suitable fault injection tools, such as ChaosMonkey and Gremlin, which provide a rich variety of fault types (such as network latency, disk I / O errors, memory leaks, etc.).

[0085] Configure the fault injection strategy, including fault type, fault frequency, fault scope, and recovery mechanism.

[0086] Design specific test cases according to the high-priority key operation scenarios, for example: High-frequency write operations: Simulate frequent write operations and evaluate their impact on the EMMC performance.

[0087] Long-duration high-load random reads: Simulate long-duration high-load random read operations and evaluate their impact on the EMMC stability and response time.

[0088] Combined operations: Simulate combinations of multiple operations (such as read immediately after write, mixed read and write, etc.) and evaluate their impact on the EMMC resource allocation and performance.

[0089] Expand the test scenarios, for example, boundary conditions, multi-user concurrency, and long-duration operation.

[0090] Further note that concentrate resources on the most critical scenarios, improve test and maintenance efficiency, and reduce unnecessary costs.

[0091] Monitoring module, execute the personalized test scenario and monitor the execution process; Execute the test cases designed according to the high-priority key operation scenarios and the test cases of the expanded test scenarios, and monitor the performance metrics during the execution process in real time.

[0092] Based on the test results, identify the performance bottlenecks of the EMMC, determine which operation scenarios have the greatest impact on performance, summarize common failure modes and their triggering conditions, and propose improvement suggestions.

[0093] According to the test results, update the fault injection tool to make it closer to real application scenarios and improve the accuracy and effectiveness of the tests.

[0094] For example, add new fault types and optimize the fault injection strategy.

[0095] The test report module generates a test report based on the execution results.

[0096] Collect detailed log files from each test node, including operation logs and error logs; Collect performance metric data from monitoring tools, such as response time, throughput, CPU usage, memory occupancy, etc.

[0097] Use a secure transmission protocol to send the test logs and performance monitoring data to the central database.

[0098] Use a visualization tool to draw trend charts of performance metrics, analyze the performance changes under different operation scenarios, identify performance bottlenecks, and determine which operation scenarios have the greatest impact on performance.

[0099] Generate a comprehensive test report that includes the fault frequency, impact scope, and recovery time.

[0100] Further illustrate that one can intuitively understand the results of this performance test, and summarize test experience and conduct the next test based on the test results.

[0101] In summary, the present invention: uses a clustering cost function for user behavior pattern recognition, accurately captures users' personalized needs and behavior habits, provides an important basis for formulating personalized test plans, improves the consistency between test results and actual user experiences, and the personalized test plan formulated thereby ensures the flexibility and adaptability of the test plan, can dynamically adjust test strategies according to different users' needs and operation scenarios, and improves the overall performance and user experience. In addition, an EMMC performance prediction model is constructed to achieve fault prediction based on multi-dimensional features, significantly improving the accuracy of fault prediction, especially for non-linear behavior prediction under high-load or abnormal operation conditions, and reducing the false alarm rate and missed alarm rate.

[0102] Embodiment 2, referring to Table 1, is the second embodiment of the present invention. To further verify the technical solution of the present invention, experimental simulation data of the performance test system of the EMMC particle controller is given.

[0103] To verify the effectiveness of the EMMC particle controller performance testing system proposed in the present invention, a series of experiments were designed and executed. The experimental object was a certain model of EMMC particle controller (named Model A), which is widely used in mobile devices to store data. The experimental environment was set under standard laboratory conditions to minimize the influence of external factors such as temperature and humidity on the experimental results.

[0104] Data collection and preprocessing, operating data: The temperature, voltage, and read / write speed data of Model A under different workloads were continuously recorded through built-in sensors. These data were recorded every 1 second and transmitted to the central database.

[0105] User historical operation record data: Data on user operation frequency, type, time, and result were collected from multiple usage scenarios, covering both normal operations and abnormal operations.

[0106] Preprocessing: After all the raw data were processed by removing noise points, filling in missing values, and normalizing, the consistency and accuracy of the data were ensured, providing a reliable basis for subsequent analysis.

[0107] Feature extraction, operating feature vector: The fast Fourier transform (FFT) method was used to extract high-dimensional features from the operating data, forming an operating feature vector containing frequency domain information.

[0108] User feature vector: The sliding window method was used to capture the number of operations of the operation frequency, and the proportions of different types of operations, time intervals, and success / failure proportions were statistically analyzed to construct a user feature vector, which reflected the user's operation habits and preferences.

[0109] Performance prediction model construction, fault mode definition: Based on the operating feature vector, multiple possible fault modes were defined, such as overheat protection activation, data loss caused by voltage fluctuations, etc.

[0110] Nonlinear mapping and combination: The operating feature transformation vector was obtained through nonlinear mapping and linearly combined with the original operating feature vector. Finally, the Sigmoid function was applied to normalize the result to generate the EMMC performance prediction value.

[0111] Fault probability calculation: The Softmax function was used to convert the scores of each fault mode into fault probabilities, and a reasonable fault threshold was set to determine whether a specific fault occurred.

[0112] User behavior pattern recognition, clustering analysis: The K-means algorithm was used to cluster the user feature vectors, obtaining several user behavior pattern clusters, and each cluster represented a typical operation style or preference.

[0113] Key operation scenario recognition and association rule mining: Based on the user behavior pattern, the FP-Growth algorithm is applied to mine frequent item sets and association rules, and the key operation scenarios affecting the EMMC performance are screened out from them.

[0114] Priority ranking: Define a weight function, calculate the priority of each operation scenario, and select the top H as the high-priority key operation scenarios.

[0115] Test plan formulation and execution, personalized test case design: For the high-priority key operation scenarios, detailed test cases are formulated and gradually extended to other scenarios to ensure comprehensive coverage.

[0116] Monitoring and logging: During the execution of the personalized test plan, various performance metrics are monitored in real-time, and all test logs and performance monitoring data are aggregated into a central database.

[0117] Test report generation, data analysis and visualization: Use data analysis tools to deeply analyze the collected data, generate charts and statistical reports, and intuitively display the performance of the EMMC particle controller under different operation scenarios. As shown in Table 1 below:

[0118] Table 1 EMMC operation comparison table

[0119] Through the comparative analysis of the above table content, it can be clearly seen that the EMMC particle controller performance test proposed by the present invention has significant advantages compared with the comparative technology: Temperature control optimization, under the same workload, the average temperature of the present invention is reduced by 5.7 °C (from 45.2 °C to 39.5 °C). This indicates that the present invention can manage heat more effectively, reduce the risk of overheating, and extend the device life.

[0120] Voltage stability improvement, the average voltage increases slightly, from 3.25V to 3.30V, indicating that the present invention can slightly improve the power supply efficiency while ensuring stable power supply, reducing the possibility of voltage fluctuations.

[0121] Read and write speed significantly improved, the average read and write speed is increased by 6.5MB / s (from 28.7MB / s to 35.2MB / s), showing the improvement of the present invention in optimizing the read and write path and greatly improving the data processing efficiency.

[0122] The failure rate is significantly reduced, the failure rate drops from 12.3% to 5.4%, reducing nearly half of the failure events. This not only proves the effectiveness of the present invention in failure prediction and prevention, but also directly improves the user experience and reliability.

[0123] The operation success rate has increased significantly, from 87.7% to 94.6%, indicating that the present invention can better adapt to different user behavior patterns and provide more stable service quality.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A performance testing system for an EMMC particle controller, characterized in that: including, a preprocessing module that collects the operation data of the EMMC particle controller and the user's historical operation record data and performs preprocessing; a feature extraction module that extracts features from the operation data and the user's historical operation record data to obtain an operation feature vector and a user feature vector; a performance prediction module that constructs an EMMC performance prediction model to obtain the failure mode and performance prediction value of the EMMC; a user analysis module that identifies the user behavior pattern according to the user feature vector; an operation scenario module that identifies the key operation scenarios based on the user behavior pattern and the failure mode; a test plan module that formulates a personalized test plan based on the key operation scenarios; a monitoring module that executes the personalized test plan and monitors the execution process; a test report module that generates a test report according to the execution result.

2. The performance test system of the EMMC particle controller according to claim 1, wherein: The operation data includes temperature data, voltage data, and read / write speed data; The user's historical operation record data includes operation frequency, operation type, operation time, and operation result.

3. The performance testing system of the EMMC particle controller according to claim 2, characterized in that: The preprocessing specifically includes the following steps: Perform noise point removal, missing value filling, and data normalization on all the collected data.

4. The performance test system of the EMMC particle controller according to claim 3, wherein: The feature extraction from the operation data and the user's historical operation record data to obtain an operation feature vector and a user feature vector specifically includes the following steps: Use the fast Fourier transform to extract the high-dimensional features of the operation data to form an operation feature vector; Use the sliding window method to capture the number of operations of the operation frequency; Statistical proportion of different operation types; Calculate the time interval between each operation; Statistical proportion of successful and failed operations; Combine the number of operations, the operation proportion, the time interval of the operation, and the proportion of successful and failed operations to form a user feature vector.

5. The performance test system of the EMMC particle controller according to claim 4, characterized in that: Construct an EMMC performance prediction model to obtain the performance prediction value of the EMMC, specifically including the following steps: Define the failure mode according to the operation feature vector; Perform a non-linear mapping on the operation feature vector to obtain an operation feature transformation vector; Perform a linear combination of the operation feature vector, the weight matrix, and the operation feature transformation vector to obtain the calculation result of the linear prediction, and apply the Sigmoid function to normalize the calculation result of the linear prediction; Perform a dot product calculation on the operation feature vector and the non-linear weight vector, and introduce a non-linear transformation to obtain a non-linear adjustment term; By combining the prediction result of the linear part and the non-linear adjustment term, obtain the failure mode score of the EMMC, and its expression is: ； Among them, represents the score of the th fault mode, represents the Sigmoid function, represents the operation feature vector, represents the transpose of the operation feature vector, represents the weight matrix of the th fault mode, represents the operation feature transformation vector, represents the weight of the non - linear adjustment term for the th fault mode, represents the non - linear weight vector; Use the Softmax function to convert the failure mode score into a failure probability; Based on the failure probability, obtain the performance prediction value of the EMMC.

6. The performance test system for the EMMC particle controller according to claim 5, wherein: The obtaining of the performance prediction value of the EMMC specifically includes the following steps: Calculate the weighted average value of the operation feature vector; Use an optimization algorithm to combine the weighted average value of the operation feature vector to calculate the failure probability to obtain the performance prediction value of the EMMC, and its expression is: ； Among them, represents the performance prediction value of the EMMC particle controller at time ; represents the number of features involved in the calculation; represents the th normalized running feature; represents the weighted average of the normalized running features; represents the square of the standard deviation of the normalized running features; represents the total number of failure modes; represents the weight of the th failure mode; represents the regularization parameter; represents the th failure mode at time failure probability.

7. The performance test system for the EMMC particle controller according to claim 6, characterized in that: The identification of the user behavior pattern according to the user feature vector specifically includes the following steps: Use the clustering cost function to select the initial cluster center from the user feature vector and calculate the Euclidean distance from each data to the cluster center; Assign the data points to the nearest cluster according to the distance from each data to the cluster center; After the clusters are allocated, calculate the center of each cluster as the mean of all data points belonging to that cluster until the clustering cost function reaches its minimum value, and its expression is: ； in, Indicates The clustering cost function at the iteration, represents the total number of clusters, Indicates The first iteration A set of data points in a cluster, represents a data point, Indicates The feature vector of the data point, Indicates The first iteration The center of the cluster; When the clustering cost function reaches its minimum value, the cluster centers and cluster assignments converge; The cluster centers are the characteristic representations of user behavior patterns, and the cluster assignments are the behavior patterns to which the user behavior characteristics belong.

8. The performance test system of the EMMC particle controller according to claim 7, characterized in that: Based on the user behavior patterns and failure probabilities, identify the key operation scenarios, which specifically include the following steps: Set the minimum support and minimum confidence; Convert the operation records of each user into a transaction format; The transaction format contains all operations within a period of time; For each user behavior pattern, apply the FP-Growth algorithm to mine frequent item sets and association rules; Use statistical analysis to find the operation scenarios that affect the performance of the EMMC particle controller from the mined association rules; Define a weight function based on the behavior patterns and failure probabilities, and calculate the priority of each operation scenario; According to the calculated priorities, sort all the key operation scenarios and select H as the high-priority key operation scenarios.

9. The performance testing system for the EMMC particle controller according to claim 8, wherein: Based on the key operation scenarios, develop a personalized test plan, which specifically includes the following steps: Build a test framework and configure a fault injection tool; Design specific test cases according to the high-priority key operation scenarios and gradually expand to other test scenarios.

10. The performance testing system of the EMMC particle controller according to claim 9, characterized in that: The generation of the test report specifically includes the following steps: Summarize all test logs and performance monitoring data into a central database; Use data analysis tools to analyze the test data and generate charts and statistical reports.

Images