A storage chip energy consumption test method and device based on multi-objective optimization

Abstract

The present invention discloses a storage chip energy consumption test method and device based on multi-objective optimization, S1, initialize the test environment; S2, set the multi-objective optimization model and fuzzy logic system; S3, construct the initial test data set; S4, and calculate the membership of each test target; S5, and generate a set of initial optimization parameter combinations; S6, based on the initial optimization parameter combination, dynamically adjust the test parameters; S7, generate the optimized test parameter combination; S8, repeat S6 and S7 until the preset optimization convergence condition is reached, and generate the final test parameter combination; S9, obtain the final test result data; S10, analyze the final test result data, comprehensively evaluate the performance of each test target through the fuzzy logic system, and generate a multi-objective optimization test report for the storage chip. The present invention effectively solves the problem of multi-objective test balance, test process optimization, and the trade-off between efficiency and accuracy existing in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of storage chips, and in particular to a storage chip energy consumption testing method and device based on multi-objective optimization. Background Art

[0002] The existing storage chip energy consumption test technology usually faces multiple challenges and limitations. The existing test methods often adopt a single-target optimization strategy, that is, they usually only focus on one aspect of the performance of the storage chip and ignore the influence of other performance indicators. Although the single-target test method obtains better test results under specific conditions, it ignores the mutual influence between different test targets, which often leads to insufficient accuracy and comprehensiveness of the test results and cannot fully reflect the comprehensive performance of the storage chip in actual use. In addition, the parameter settings in the test process in the existing technology are often static, that is, the test parameters are pre-set before the test starts, and then remain unchanged throughout the test process. Although the static parameter setting method simplifies the test process, it also greatly reduces the adaptability and accuracy of the test results. Especially in the face of complex workloads and changes in environmental conditions, the static parameter setting test method cannot adjust the test strategy in real time, so it is impossible to obtain the optimal test results, which makes the test results have limited reference value in practical applications.

[0003] In addition, the existing storage chip energy consumption test technology still has a trade-off problem between test efficiency and accuracy. In order to obtain comprehensive test data, it is usually necessary to test under multiple test conditions. Although the test method can improve the accuracy of the test results, it also brings about the problems of long test time, large resource consumption, and high test cost. Especially in the face of large-scale production, the time-consuming test method is particularly unsuitable, resulting in reduced production efficiency. In the prior art, although some methods try to improve the test efficiency by optimizing the test process or simplifying the test steps, the accuracy of the test results is often sacrificed in the process, which makes it difficult to obtain comprehensive and accurate test data in practical applications, further limiting the optimization and improvement of storage chip performance.

[0004] In summary, the main defects of the existing storage chip energy consumption testing technology are that it is difficult to take into account multiple test objectives such as energy consumption, performance and stability at the same time, and it is impossible to achieve dynamic parameter adjustment during the test process, making it difficult to obtain comprehensive and accurate test results. At the same time, the trade-off between test efficiency and accuracy has not been effectively resolved. The existence of these problems has seriously restricted the design and optimization process of storage chips and affected their performance in practical applications. Therefore, there is an urgent need for a storage chip energy consumption testing method and device based on multi-objective optimization to solve the above-mentioned defects in the prior art. Summary of the invention

[0005] One purpose of the present invention is to propose a storage chip energy consumption testing method and device based on multi-objective optimization. The present invention effectively solves the problems of multi-objective test balance, test process optimization and the trade-off between efficiency and accuracy existing in the prior art.

[0006] A storage chip energy consumption test method and device based on multi-objective optimization according to an embodiment of the present invention comprises the following steps:

[0007] S1. Initialize the test environment, connect the storage chip and set the initial test parameters, including voltage, current, frequency and temperature;

[0008] S2. Setting a multi-objective optimization model and a fuzzy logic system to determine multiple test targets to be optimized, the test targets including energy consumption, performance and stability of the storage chip;

[0009] S3. Run the test program to obtain the energy consumption, performance and stability indicators of the storage chip under different voltage, current, frequency and temperature conditions, and build an initial test data set;

[0010] S4, using the fuzzy logic system to process the initial test data set, converting the test target into a fuzzy set according to the fuzzy rules, and calculating the membership degree of each test target;

[0011] S5. According to the membership degree output by the fuzzy logic system, the optimization weight of each test target is calculated through the multi-objective optimization model, and a set of initial optimization parameter combinations is generated;

[0012] S6. Based on the initial optimization parameter combination, dynamically adjust the test parameters, obtain the adjusted test data in real time, and re-evaluate the membership of the test target through the fuzzy logic system;

[0013] S7, updating the optimization weights in the multi-objective optimization model according to the membership degree re-evaluated by the fuzzy logic system, and generating an optimized test parameter combination;

[0014] S8, repeat S6 and S7 until the preset optimization convergence condition is reached, and generate the final test parameter combination;

[0015] S9, using the final test parameter combination, performing a full range of energy consumption tests on the storage chip to obtain the final test result data;

[0016] S10. Analyze the final test result data, comprehensively evaluate the performance of each test target through the fuzzy logic system, and generate a multi-target optimization test report for the storage chip.

[0017] Optionally, S3 specifically includes the following steps:

[0018] S31, initialize the test environment, connect the storage chip, put the storage chip into a test state, and set initial test parameters, the initial test parameters include voltage V0, current I0, frequency f0 and temperature T0;

[0019] S32, running the test program, performing the test under the initial test parameters, and obtaining the energy consumption data P0, performance data Q0 and stability data S0 of the storage chip;

[0020] S33, repeat step S32, at different voltages V i 、Current I i , frequency f i and temperature T i Under these conditions, the energy consumption data P corresponding to the storage chip is obtained respectively. i 、Performance dataQ i and stability data S i , where i = 1, 2, 3, ..., n, n is the number of tests;

[0021] S34, the energy consumption data P i 、Performance dataQ i and stability data S i Perform normalization to form a normalized data set {P′ i ,Q′ i ,S′ i}, where P′ i , Q′ i and S′ i They represent the normalized energy consumption, performance, and stability indicator data sets under different test conditions;

[0022] S35, construct an initial test data set D0, including all normalized energy consumption data P′ i , performance data Q′ i and stability data S′ i :

[0023] D0={P′ i ,Q′ i ,S′ i |i=1,2,3,…,n}.

[0024] Optionally, the S4 specifically includes the following steps:

[0025] S41, constructing a fuzzy logic system, wherein the fuzzy logic system comprises:

[0026] Define the input variables of the fuzzy set. Based on the energy consumption test target of the storage chip, determine the input variables including energy consumption P, performance Q and stability S. Fuzzify the input variables and define them as fuzzy sets F X, where X is energy consumption P, performance Q and stability S;

[0027] Construct a fuzzy rule base and formulate fuzzy rules based on the empirical data and technical requirements of storage chip energy consumption testing:

[0028] R i : If P is A i And Q is B i And S is C i , then i =min(μ P (x),μ Q (y),μ S (z));

[0029] Among them, R i represents the i-th fuzzy rule, A i , B i and C i are fuzzy subsets of energy consumption, performance and stability, respectively, i Output the membership value under the rule. The fuzzy rule base contains multiple similar rules to cover various possibilities of storage chip testing;

[0030] Construct the membership function. For each fuzzy subset, define the membership function as follows:

[0031]

[0032] in, is the membership value of the input variable X in different fuzzy subsets, is the weight coefficient of the fuzzy subset, which is obtained by weighted averaging the fuzzified data;

[0033] S42, input the initial test data set D0 into the fuzzification module of the constructed fuzzy logic system, and transform each test target into the corresponding fuzzy set F according to the rules in the fuzzy rule base. X , and calculate the membership:

[0034]

[0035] Among them, t is the corresponding input variable, is the membership value;

[0036] S43, in the membership calculation module, according to the fuzzy rules in the fuzzy rule base, calculate the fuzzy set F P 、F Q 、F S The membership value of each element in is obtained to obtain the membership matrix in They represent the test target P′ respectively.i , Q′ i , S′ i Membership values ​​under different test conditions:

[0037]

[0038] Optionally, the S5 specifically includes the following steps:

[0039] S51, input the membership matrix into the multi-objective optimization model, the multi-objective optimization model is constructed based on the particle swarm optimization algorithm, and initialize the speed of each particle and location

[0040]

[0041] in, Represents the position vector of the jth particle, including the initial optimization weight is the velocity vector, η1 and ζ1 are learning factors, and p best is the best historical position of the individual particle, g best is the global optimal position;

[0042] S52, by optimizing the fitness function F(ω P ,ω Q ,ω S ) Calculate the fitness value of each particle:

[0043]

[0044] Among them, p j is the particle's position vector, σ P , σ Q , σ S are the standard deviations of energy consumption, performance and stability, μ P , μ Q , μ S is the corresponding mean, β1 is the penalty factor, which is used to reflect the stability of the target in the fitness function;

[0045] S53, through the iterative calculation of the particle swarm optimization algorithm, the position and speed of the particle are updated until the fitness function F(ω P ,ω Q ,ω S ) to achieve the best result and generate a set of initial optimization parameter combinations

[0046]

[0047] Where w is the inertia weight, is the velocity vector after the t+1th iteration, is the particle position after the t+1th iteration.

[0048] Optionally, the S6 specifically includes the following steps:

[0049] S61, combining the initial optimization parameters Applied to the test process to set the corresponding test parameters in, Respectively represent the optimization weights Adjusted voltage, current, frequency and temperature parameters;

[0050] S62. Run the test program under the dynamically adjusted test parameters to obtain the energy consumption data of the storage chip in real time. Performance Data and stability data

[0051] S63, dynamically adjusted test data Input the fuzzy logic system, re-evaluate the membership of each test target according to the constructed fuzzy rule base, and calculate the membership matrix:

[0052]

[0053] in, They represent the membership values ​​of energy consumption, performance and stability under the dynamically adjusted test conditions respectively;

[0054] S64, according to the re-evaluated membership matrix Λ * , for the optimization parameter combination Further adjustments are made to form a new test parameter combination, and steps S61-S63 are repeated until the membership matrix Λ * Converge or reach the preset optimization criteria.

[0055] Optionally, the S7 specifically includes the following contents:

[0056] S71, input the membership matrix obtained in S6 into the multi-objective optimization model, and recalculate the optimization weights of each test target

[0057]

[0058] in, represents the optimization weight of the tth iteration, α1 is the learning rate, γ1 is the adjustment parameter in the model, is the membership degree after dynamic adjustment;

[0059] S72, in the process of updating the optimization weight, using the fitness function Evaluate and adjust optimization weights:

[0060]

[0061] Among them, δ1 is the penalty factor, which is used to reflect the volatility of the target in the fitness function, σ P , σ Q , σ S are the standard deviations of energy consumption, performance and stability, μ P , μ Q , μ S is the corresponding mean;

[0062] S73. Based on the fitness function evaluation, the optimization weight is adjusted through an iterative optimization algorithm until the fitness function converges to the optimal value:

[0063]

[0064] in, is the weight vector of the tth iteration, η2 is the learning rate, ξ2 is the momentum factor, is the gradient of the fitness function.

[0065] Optionally, the S9 specifically includes the following steps:

[0066] S91. Apply the final optimized test parameter combination to the full range energy consumption test of the storage chip, and set the test parameters including the final voltage Current frequency and temperature

[0067] S92. Under the final optimized test parameter combination, run the full range energy consumption test program to obtain the energy consumption data P of the storage chip under different working modes and load conditions. final 、Performance dataQ final and stability data S final ;

[0068] S93, the energy consumption data P final 、Performance dataQ final and stability data S final Perform classification and normalization to form the final test data set {P′ final ,Q′ final ,S′ final};

[0069] S94. Based on the final test data set, calculate the comprehensive energy consumption performance index C of the storage chip final :

[0070]

[0071] Among them, μ(P′ final ), μ(Q′ final ) and μ(S′ final ) represent the mean values ​​of energy consumption, performance and stability indicators respectively, and is the final optimization weight;

[0072] S95, the comprehensive energy consumption performance index C final Compare with the preset target value and generate the final test result report of the storage chip:

[0073]

[0074] Among them, C target is the preset target value, and sign(·) is the sign function used to determine the positivity of the test result.

[0075] A storage chip energy consumption test device based on multi-objective optimization includes the following modules:

[0076] The test environment initialization module is used to initialize the test environment, connect the storage chip and set the initial test parameters;

[0077] The multi-objective optimization model module integrates the fuzzy logic system and the multi-objective optimization model to build a fuzzy rule base based on the input test data set, generate and adjust the optimization weights, and output the optimized test parameter combination;

[0078] A test execution module, connected to the test environment initialization module, is used to perform a full range of energy consumption tests according to the optimized test parameter combination, and obtain energy consumption, performance and stability data of the storage chip under different working modes and load conditions;

[0079] A data processing module is used to classify and normalize the energy consumption, performance and stability data obtained by the test execution module to generate a final test data set;

[0080] The comprehensive energy consumption performance analysis module is used to calculate the comprehensive energy consumption performance index of the storage chip based on the final test data set, and compare the comprehensive energy consumption performance index with the preset target value to generate the final test result report of the storage chip;

[0081] The display module is used to display the final test result report and store the real-time data and analysis results of the chip during the test process.

[0082] The beneficial effects of the present invention are:

[0083] (1) The present invention improves the comprehensiveness and accuracy of the test results by introducing a multi-objective optimization algorithm. In the prior art, the test only focuses on a single objective and it is difficult to take into account multiple indicators such as energy consumption, performance and stability. The present invention constructs a multi-objective optimization model based on a fuzzy logic system. During the test process, multiple conflicting test objectives can be optimized simultaneously. By dynamically adjusting the optimization weights of each objective, the optimal test parameter combination can be automatically found under different test conditions, so that the test results not only reflect the best performance of a single indicator, but also achieve the best balance in comprehensive performance.

[0084] (2) The present invention realizes real-time optimization during the test process by designing a dynamic parameter adjustment mechanism. By obtaining test feedback in real time, the fuzzy logic system is used to instantly evaluate the test data and dynamically adjust the test parameters according to the evaluation results. The present invention can adapt to different workloads and environmental conditions during the test process, so that the test system can always maintain the best state under different test conditions, thereby improving the accuracy and reliability of the test results.

[0085] (3) The present invention achieves an effective balance between test efficiency and accuracy. By combining the optimization algorithm with the data processing module, the test efficiency can be significantly improved while ensuring the accuracy of the test results. The iterative calculation of the test parameter combination by the optimization algorithm reduces unnecessary test steps. The data processing process is simplified through normalization processing and fuzzy logic evaluation, thereby shortening the test time and reducing the test cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0087] Figure 1 The present invention provides a flow chart of a storage chip energy consumption testing method and device based on multi-objective optimization. DETAILED DESCRIPTION

[0088] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0089] refer to Figure 1 , a storage chip energy consumption test method and device based on multi-objective optimization, comprising the following steps:

[0090] S1. Initialize the test environment, connect the storage chip and set the initial test parameters, including voltage, current, frequency and temperature;

[0091] S2. Setting a multi-objective optimization model and a fuzzy logic system to determine multiple test targets to be optimized, the test targets including energy consumption, performance and stability of the storage chip;

[0092] S3. Run the test program to obtain the energy consumption, performance and stability indicators of the storage chip under different voltage, current, frequency and temperature conditions, and build an initial test data set;

[0093] S4, using the fuzzy logic system to process the initial test data set, converting the test target into a fuzzy set according to the fuzzy rules, and calculating the membership degree of each test target;

[0094] S5. According to the membership degree output by the fuzzy logic system, the optimization weight of each test target is calculated through the multi-objective optimization model, and a set of initial optimization parameter combinations is generated;

[0095] S6. Based on the initial optimization parameter combination, dynamically adjust the test parameters, obtain the adjusted test data in real time, and re-evaluate the membership of the test target through the fuzzy logic system;

[0096] S7, updating the optimization weights in the multi-objective optimization model according to the membership degree re-evaluated by the fuzzy logic system, and generating an optimized test parameter combination;

[0097] S8, repeat S6 and S7 until the preset optimization convergence condition is reached, and generate the final test parameter combination;

[0098] S9, using the final test parameter combination, performing a full range of energy consumption tests on the storage chip to obtain the final test result data;

[0099] S10. Analyze the final test result data, comprehensively evaluate the performance of each test target through the fuzzy logic system, and generate a multi-target optimization test report for the storage chip.

[0100] In this implementation, S3 specifically includes the following steps:

[0101] S31, initialize the test environment, connect the storage chip, put the storage chip into a test state, and set initial test parameters, the initial test parameters include voltage V0, current I0, frequency f0 and temperature T0;

[0102] S32, running the test program, performing the test under the initial test parameters, and obtaining the energy consumption data P0, performance data Q0 and stability data S0 of the storage chip;

[0103] S33, repeat step S32, at different voltages V i 、Current I i , frequency f i and temperature Ti Under these conditions, the energy consumption data P corresponding to the storage chip is obtained respectively. i 、Performance dataQ i and stability data S i , where i = 1, 2, 3, ..., n, n is the number of tests;

[0104] S34, the energy consumption data P i 、Performance dataQ i and stability data S i Normalization is performed to form a normalized data set {P′ i ,Q′ i ,S′ i}, where P′ i , Q′ i and S′ i They represent the normalized energy consumption, performance, and stability indicator data sets under different test conditions;

[0105] S35, construct an initial test data set D0, including all normalized energy consumption data P′ i , performance data Q′ i and stability data S′ i :

[0106] D0={P′ i ,Q′ i ,S′ i |i=1,2,3,…,n}.

[0107] In this implementation, S4 specifically includes the following steps:

[0108] S41. Construct a fuzzy logic system. The fuzzy logic system includes:

[0109] Define the input variables of the fuzzy set. Based on the energy consumption test target of the storage chip, determine the input variables including energy consumption P, performance Q and stability S. Fuzzify the input variables and define them as fuzzy sets F X , where X is energy consumption P, performance Q and stability S;

[0110] Construct a fuzzy rule base and formulate fuzzy rules based on the empirical data and technical requirements of storage chip energy consumption testing:

[0111] R i : If P is A i And Q is B i And S is C i , then i =min(μ P (x),μ Q (y),μ S (z));

[0112] Among them, R i represents the i-th fuzzy rule, A i , B i and C i are fuzzy subsets of energy consumption, performance and stability, respectively, i Output the membership value under the rule. The fuzzy rule base contains multiple similar rules to cover various possibilities of storage chip testing;

[0113] Construct the membership function. For each fuzzy subset, define the membership function as follows:

[0114]

[0115] in, is the membership value of the input variable X in different fuzzy subsets, is the weight coefficient of the fuzzy subset, which is obtained by weighted averaging the fuzzified data;

[0116] S42, input the initial test data set D0 into the fuzzification module of the constructed fuzzy logic system, and transform each test target into the corresponding fuzzy set F according to the rules in the fuzzy rule base. X , and calculate the membership:

[0117]

[0118] Among them, t is the corresponding input variable, is the membership value;

[0119] S43, in the membership calculation module, according to the fuzzy rules in the fuzzy rule base, calculate the fuzzy set F P 、F Q 、F S The membership value of each element in is obtained to obtain the membership matrix in They represent the test target P′ respectively. i , Q′ i , S′ i Membership values ​​under different test conditions:

[0120]

[0121] In this implementation, S5 specifically includes the following steps:

[0122] S51, input the membership matrix into the multi-objective optimization model, the multi-objective optimization model is constructed based on the particle swarm optimization algorithm, and initialize the speed of each particle and location

[0123]

[0124] in, Represents the position vector of the jth particle, including the initial optimization weight is the velocity vector, η1 and ζ1 are learning factors, and p best is the best historical position of the individual particle, g best is the global optimal position;

[0125] S52, by optimizing the fitness function F(ω P ,ω Q ,ω S ) Calculate the fitness value of each particle:

[0126]

[0127] Among them, p j is the particle's position vector, σ P , σ Q , σ S are the standard deviations of energy consumption, performance and stability, μ P , μ Q , μ S is the corresponding mean, β1 is the penalty factor, which is used to reflect the stability of the target in the fitness function;

[0128] S53, through the iterative calculation of the particle swarm optimization algorithm, the position and speed of the particle are updated until the fitness function F(ω P ,ω Q ,ω S ) to achieve the best result and generate a set of initial optimization parameter combinations

[0129]

[0130] Where w is the inertia weight, is the velocity vector after the t+1th iteration, is the particle position after the t+1th iteration.

[0131] In this implementation, S6 specifically includes the following steps:

[0132] S61, combining the initial optimization parameters Applied to the test process to set the corresponding test parameters in, Respectively represent the optimization weights Adjusted voltage, current, frequency and temperature parameters;

[0133] S62. Run the test program under the dynamically adjusted test parameters to obtain the energy consumption data of the storage chip in real time. Performance Data and stability data

[0134] S63, dynamically adjusted test data Input the fuzzy logic system, re-evaluate the membership of each test target according to the constructed fuzzy rule base, and calculate the membership matrix:

[0135]

[0136] in, They represent the membership values ​​of energy consumption, performance and stability under the dynamically adjusted test conditions respectively;

[0137] S64, according to the re-evaluated membership matrix Λ * , for the optimization parameter combination Further adjustments are made to form a new test parameter combination, and steps S61-S63 are repeated until the membership matrix Λ * Converge or reach the preset optimization criteria.

[0138] In this implementation, S7 specifically includes the following contents:

[0139] S71, input the membership matrix obtained in S6 into the multi-objective optimization model, and recalculate the optimization weights of each test target

[0140]

[0141] in, represents the optimization weight of the tth iteration, α1 is the learning rate, γ1 is the adjustment parameter in the model, is the membership degree after dynamic adjustment;

[0142] S72, in the process of updating the optimization weight, using the fitness function Evaluate and adjust optimization weights:

[0143]

[0144] Among them, δ1 is the penalty factor, which is used to reflect the volatility of the target in the fitness function, σ P , σ Q , σ S are the standard deviations of energy consumption, performance and stability, μ P , μ Q , μ S is the corresponding mean;

[0145] S73. Based on the fitness function evaluation, the optimization weight is adjusted through an iterative optimization algorithm until the fitness function converges to the optimal value:

[0146]

[0147] in, is the weight vector of the tth iteration, η2 is the learning rate, ξ2 is the momentum factor, is the gradient of the fitness function.

[0148] In this implementation, S9 specifically includes the following steps:

[0149] S91. Apply the final optimized test parameter combination to the full range energy consumption test of the storage chip, and set the test parameters including the final voltage Current frequency and temperature

[0150] S92. Under the final optimized test parameter combination, run the full range energy consumption test program to obtain the energy consumption data P of the storage chip under different working modes and load conditions. final 、Performance dataQ final and stability data S final ;

[0151] S93, the energy consumption data P final 、Performance dataQ final and stability data S final Perform classification and normalization to form the final test data set {P′ final ,Q′ final ,S′ final};

[0152] S94. Based on the final test data set, calculate the comprehensive energy consumption performance index C of the storage chip final :

[0153]

[0154] Among them, μ(P′ final ), μ(Q′ final ) and μ(S′ final ) represent the mean values ​​of energy consumption, performance and stability indicators respectively, and is the final optimization weight;

[0155] S95, the comprehensive energy consumption performance index C final Compare with the preset target value and generate the final test result report of the storage chip:

[0156]

[0157] Among them, C target is the preset target value, and sign(·) is the sign function used to determine the positivity of the test result.

[0158] A storage chip energy consumption test device based on multi-objective optimization includes the following modules:

[0159] The test environment initialization module is used to initialize the test environment, connect the storage chip and set the initial test parameters;

[0160] The multi-objective optimization model module integrates the fuzzy logic system and the multi-objective optimization model to build a fuzzy rule base based on the input test data set, generate and adjust the optimization weights, and output the optimized test parameter combination;

[0161] A test execution module, connected to the test environment initialization module, is used to perform a full range of energy consumption tests according to the optimized test parameter combination, and obtain energy consumption, performance and stability data of the storage chip under different working modes and load conditions;

[0162] A data processing module is used to classify and normalize the energy consumption, performance and stability data obtained by the test execution module to generate a final test data set;

[0163] The comprehensive energy consumption performance analysis module is used to calculate the comprehensive energy consumption performance index of the storage chip based on the final test data set, and compare the comprehensive energy consumption performance index with the preset target value to generate the final test result report of the storage chip;

[0164] The display module is used to display the final test result report and store the real-time data and analysis results of the chip during the test process.

[0165] Embodiment 1:

[0166] In actual applications, a well-known semiconductor manufacturing company is developing a new low-power storage chip for the high-end smartphone market. The goal is to ensure that the chip achieves the best balance between energy consumption, performance and stability to meet the growing market demand. In the early tests, the company adopted the traditional single-target optimization test method, but in the comprehensive performance test, it was found that although the power consumption of the chip was reduced, the data transmission rate and stability decreased to varying degrees. Especially in high temperature environments, the error rate of the chip increased significantly, seriously affecting the overall performance. In order to solve this problem, the company decided to introduce the present invention to further optimize the chip design and improve the test efficiency.

[0167] In July 2023, the company selected a 128GB storage chip under development for testing. First, the testing team initialized the testing environment using the testing device of the present invention and set the initial testing parameters: voltage of 3.3V, current of 500mA, operating frequency of 200MHz, and temperature controlled at 25°C. These parameters were set based on the optimal initial conditions obtained in previous experiments and were intended to simulate the working state of the chip in normal use.

[0168] In the first phase of the test, the system ran a preliminary full-range test. Data records showed that under the conditions of frequency 200MHz and voltage 3.3V, the chip's average power consumption was 140mW, the data transmission rate was 105MB / s, and the error rate was 0.05%. Subsequently, based on these preliminary data, the system used fuzzy logic system and multi-objective optimization algorithm to make preliminary weight allocations for the three goals of energy consumption, performance and stability, and generated the first set of optimized test parameter combinations: voltage 3.2V, frequency 195MHz, and temperature remained unchanged.

[0169] In the second phase of testing, the system was further tested according to the optimized parameters. Test data showed that the average power consumption of the chip dropped to 135mW, the data transmission rate increased to 108MB / s, and the error rate decreased to 0.04%. The system again calculated a new membership matrix through the fuzzy logic system and adjusted the weights of each test target to generate the second set of optimized parameter combinations: voltage 3.15V, frequency 190MHz, and temperature increased to 30°C. At this time, the team noticed that at higher temperatures, although the chip's energy consumption and performance were still good, the error rate had a slight upward trend, reaching 0.045%.

[0170] In the third phase of testing, the system further adjusted the test parameters, lowered the voltage to 3.1V, adjusted the frequency to 185MHz, and controlled the temperature between 25℃ and 30℃. The test results showed that the average power consumption of the chip dropped to 130mW, the data transmission rate stabilized at 110MB / s, and the error rate dropped to 0.03%. After this round of testing, the system calculated the final membership matrix and generated the final optimized parameter combination: voltage 3.05V, frequency 180MHz, and temperature maintained at 25℃.

[0171] Through the method of the present invention, the testing team recorded detailed data during the testing process (Table 1), which shows the significant difference between the method of the present invention and the traditional testing method:

[0172] Table 1 Comparison data between the method of the present invention and the traditional test method

[0173]

[0174] As can be seen from Table 1 above, during the entire test process, the average power consumption of the chip is gradually optimized from the initial 140mW to 127mW, which is about 9% lower than that of the traditional method. At the same time, the data transmission rate is increased from 105MB / s to 112MB / s, and the stability (expressed as the error rate) is reduced from 0.05% to 0.02%. The test time is shortened from 4 hours of the traditional method to 1.5 hours.

[0175] At the test summary meeting in August 2023, the company conducted a detailed analysis of the test results and finally confirmed that the method of the present invention not only significantly optimized the energy consumption and performance of the storage chip, but also reduced unnecessary testing steps and time expenses through multiple rounds of dynamic optimization, ensuring that the chip's comprehensive performance is better than traditional testing methods. Based on this test result, the company decided to fully apply the method of the present invention to the development and testing process of the next-generation storage chip, providing strong technical support for future product competitiveness.

[0176] This embodiment fully verifies the feasibility and effectiveness of the present invention through specific test data and optimization process, demonstrates its broad application prospects in storage chip energy consumption testing, effectively solves the problems of insufficient multi-objective optimization, difficult dynamic adjustment and low efficiency in traditional testing methods, and provides new technical ideas for the innovative development of the semiconductor industry.

[0177] The present invention improves the comprehensiveness and accuracy of the test results by introducing a multi-objective optimization algorithm. In the prior art, the test only focuses on a single goal and it is difficult to take into account multiple indicators such as energy consumption, performance and stability. The present invention constructs a multi-objective optimization model based on a fuzzy logic system, which can simultaneously optimize multiple conflicting test objectives during the test process. By dynamically adjusting the optimization weights of each objective, the optimal test parameter combination is automatically found under different test conditions, so that the test results not only reflect the best performance of a single indicator, but also can achieve the best balance in comprehensive performance.

[0178] The present invention realizes real-time optimization in the test process by designing a dynamic parameter adjustment mechanism. By obtaining test feedback in real time, the fuzzy logic system is used to instantly evaluate the test data, and the test parameters are dynamically adjusted according to the evaluation results. It can adapt to different workloads and environmental conditions during the test process, so that the test system can always maintain the best state under different test conditions, thereby improving the accuracy and reliability of the test results.

[0179] The present invention achieves an effective balance between test efficiency and accuracy. By combining the optimization algorithm with the data processing module, the test efficiency can be significantly improved while ensuring the accuracy of the test results. The iterative calculation of the test parameter combination by the optimization algorithm reduces unnecessary test steps. The data processing flow is simplified through normalization processing and fuzzy logic evaluation, thereby shortening the test time and reducing the test cost.

[0180] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A storage chip energy consumption test method based on multi-objective optimization, characterized in that: The steps include: S1. Initialize the test environment, connect the storage chip and set the initial test parameters, including voltage, current, frequency and temperature; S2. Setting a multi-objective optimization model and a fuzzy logic system to determine multiple test targets to be optimized, the test targets including energy consumption, performance and stability of the storage chip; S3. Run the test program to obtain the energy consumption, performance and stability indicators of the storage chip under different voltage, current, frequency and temperature conditions, and build an initial test data set; S4, using the fuzzy logic system to process the initial test data set, converting the test target into a fuzzy set according to the fuzzy rules, and calculating the membership degree of each test target; S5. According to the membership degree output by the fuzzy logic system, the optimization weight of each test target is calculated through the multi-objective optimization model, and a set of initial optimization parameter combinations is generated; S6. Based on the initial optimization parameter combination, dynamically adjust the test parameters, obtain the adjusted test data in real time, and re-evaluate the membership of the test target through the fuzzy logic system; S7, updating the optimization weights in the multi-objective optimization model according to the membership degree re-evaluated by the fuzzy logic system, and generating an optimized test parameter combination; S8, repeat S6 and S7 until the preset optimization convergence condition is reached, and generate the final test parameter combination; S9, using the final test parameter combination, performing a full range of energy consumption tests on the storage chip to obtain the final test result data; S10. Analyze the final test result data, comprehensively evaluate the performance of each test target through the fuzzy logic system, and generate a multi-target optimization test report for the storage chip.

2. The storage chip energy consumption test method based on multi-objective optimization according to claim 1 is characterized in that: The S3 specifically includes the following steps: S31, initialize the test environment, connect the storage chip, put the storage chip into a test state, and set initial test parameters, the initial test parameters include voltage V0, current I0, frequency f0 and temperature T0; S32, running the test program, performing the test under the initial test parameters, and obtaining the energy consumption data P0, performance data Q0 and stability data S0 of the storage chip; S33, repeat step S32, at different voltages V i 、Current I i , frequency f i and temperature T i Under these conditions, the energy consumption data P corresponding to the storage chip is obtained respectively. i 、Performance dataQ i and stability data S i , where i = 1, 2, 3, ..., n, n is the number of tests; S34, the energy consumption data P i 、Performance dataQ i and stability data S i Perform normalization to form a normalized data set {P′ i ,Q′ i ,S′ i }, where P′ i , Q′ i and S′ i They represent the normalized energy consumption, performance, and stability indicator data sets under different test conditions; S35, construct an initial test data set D0, including all normalized energy consumption data P′ i , performance data Q′ i and stability data S′ i : D0={P′ i ,Q′ i ,S′ i |i=1,2,3,…,n}。 3. The storage chip energy consumption test method based on multi-objective optimization according to claim 1 is characterized in that: The S4 specifically comprises the following steps: S41, constructing a fuzzy logic system, wherein the fuzzy logic system comprises: Define the input variables of the fuzzy set. Based on the energy consumption test target of the storage chip, determine the input variables including energy consumption P, performance Q and stability S. Fuzzify the input variables and define them as fuzzy sets F X , where X is energy consumption P, performance Q and stability S; Construct a fuzzy rule base and formulate fuzzy rules based on the empirical data and technical requirements of storage chip energy consumption testing: R i : If P is A i And Q is B i And S is C i , then i =min(μ P (x),μ Q (y),μ S (z)); Among them, R i represents the i-th fuzzy rule, A i , B i and C i are fuzzy subsets of energy consumption, performance and stability, respectively, i The membership output value under the rule is the fuzzy rule base, which contains multiple similar rules to cover various possibilities of storage chip testing; Construct the membership function. For each fuzzy subset, define the membership function as follows: in, is the membership value of the input variable X in different fuzzy subsets, is the weight coefficient of the fuzzy subset, which is obtained by weighted averaging the fuzzified data; S42, input the initial test data set D0 into the fuzzification module of the constructed fuzzy logic system, and transform each test target into the corresponding fuzzy set F according to the rules in the fuzzy rule base. X , and calculate the membership: Among them, t is the corresponding input variable, is the membership value; S43, in the membership calculation module, according to the fuzzy rules in the fuzzy rule base, calculate the fuzzy set F P 、F Q 、F S The membership value of each element in is obtained to obtain the membership matrix in, They represent the test target P′ respectively. i , Q′ i , S′ i Membership values ​​under different test conditions.

4. The storage chip energy consumption test method based on multi-objective optimization according to claim 1, characterized in that: The S5 specifically includes the following steps: S51, input the membership matrix into the multi-objective optimization model, the multi-objective optimization model is constructed based on the particle swarm optimization algorithm, and initialize the speed of each particle and location in, Represents the position vector of the jth particle, including the initial optimization weight is the velocity vector, η1 and ζ1 are learning factors, and p best is the best historical position of the individual particle, g best is the global optimal position; S52, by optimizing the fitness function F(ω P ,ω Q ,ω S ) Calculate the fitness value of each particle: Among them, p j is the particle's position vector, σ P , σ Q , σ S are the standard deviations of energy consumption, performance and stability, μ P , μ Q , μ S is the corresponding mean, β1 is the penalty factor, which is used to reflect the stability of the target in the fitness function; S53, through the iterative calculation of the particle swarm optimization algorithm, the position and speed of the particle are updated until the fitness function F(ω P ,ω Q ,ω S ) to achieve the best result and generate a set of initial optimization parameter combinations Where w is the inertia weight, is the velocity vector after the t+1th iteration, is the particle position after the t+1th iteration.

5. The storage chip energy consumption test method based on multi-objective optimization according to claim 1, characterized in that: The S6 specifically comprises the following steps: S61, combining the initial optimization parameters Applied to the test process to set the corresponding test parameters in, Respectively represent the optimization weights Adjusted voltage, current, frequency and temperature parameters; S62. Run the test program under the dynamically adjusted test parameters to obtain the energy consumption data of the storage chip in real time. Performance Data and stability data S63, dynamically adjusted test data Input the fuzzy logic system, re-evaluate the membership of each test target according to the constructed fuzzy rule base, and calculate the membership matrix: in, They represent the membership values ​​of energy consumption, performance and stability under the dynamically adjusted test conditions respectively; S64, according to the re-evaluated membership matrix Λ * , for the optimization parameter combination Further adjustments are made to form a new test parameter combination, and steps S61-S63 are repeated until the membership matrix Λ * Converge or reach the preset optimization criteria.

6. The storage chip energy consumption test method based on multi-objective optimization according to claim 1, characterized in that: The S7 specifically includes the following contents: S71, input the membership matrix obtained in S6 into the multi-objective optimization model, and recalculate the optimization weights of each test target in, represents the optimization weight of the tth iteration, α1 is the learning rate, γ1 is the adjustment parameter in the model, is the membership degree after dynamic adjustment; S72, in the process of updating the optimization weight, using the fitness function Evaluate and adjust optimization weights: Among them, δ1 is the penalty factor, which is used to reflect the volatility of the target in the fitness function, σ P , σ Q , σ S are the standard deviations of energy consumption, performance and stability, μ P , μ Q , μ S is the corresponding mean; S73. Based on the fitness function evaluation, the optimization weight is adjusted through an iterative optimization algorithm until the fitness function converges to the optimal value: in, is the weight vector of the tth iteration, η2 is the learning rate, ξ2 is the momentum factor, is the gradient of the fitness function.

7. The storage chip energy consumption test method based on multi-objective optimization according to claim 1, characterized in that: The S9 specifically includes the following steps: S91. Apply the final optimized test parameter combination to the full range energy consumption test of the storage chip, and set the test parameters including the final voltage Current frequency and temperature S92. Under the final optimized test parameter combination, run the full range energy consumption test program to obtain the energy consumption data P of the storage chip under different working modes and load conditions. final 、Performance dataQ final and stability data S final ; S93, the energy consumption data P final 、Performance dataQ final and stability data S final Classify and Normalization processing to form the final test data set {P final ,Q final ,S final }; S94. Based on the final test data set, calculate the comprehensive energy consumption performance index C of the storage chip final : ″′ Among them, μ(P final )、μ(Q final ) and μ(S final ) represent the mean values ​​of energy consumption, performance and stability indicators respectively, and is the final optimization weight; S95, the comprehensive energy consumption performance index C final Compare with the preset target value and generate the final test result report of the storage chip: Among them, C target is the preset target value, and sign(·) is the sign function used to determine the positivity of the test result.

8. A storage chip energy consumption test device based on multi-objective optimization, characterized in that: Includes the following modules: The test environment initialization module is used to initialize the test environment, connect the storage chip and set the initial test parameters; The multi-objective optimization model module integrates the fuzzy logic system and the multi-objective optimization model to build a fuzzy rule base based on the input test data set, generate and adjust the optimization weights, and output the optimized test parameter combination; A test execution module, connected to the test environment initialization module, is used to perform a full range of energy consumption tests according to the optimized test parameter combination, and obtain energy consumption, performance and stability data of the storage chip under different working modes and load conditions; A data processing module is used to classify and normalize the energy consumption, performance and stability data obtained by the test execution module to generate a final test data set; The comprehensive energy consumption performance analysis module is used to calculate the comprehensive energy consumption performance index of the storage chip based on the final test data set, and compare the comprehensive energy consumption performance index with the preset target value to generate the final test result report of the storage chip; The display module is used to display the final test result report and store the real-time data and analysis results of the chip during the test process.

Images