WO2022062567A1 - 产品测试数据的检测方法、系统、电子设备和存储介质 - Google Patents

产品测试数据的检测方法、系统、电子设备和存储介质 Download PDF

Info

Publication number
WO2022062567A1
WO2022062567A1 PCT/CN2021/104882 CN2021104882W WO2022062567A1 WO 2022062567 A1 WO2022062567 A1 WO 2022062567A1 CN 2021104882 W CN2021104882 W CN 2021104882W WO 2022062567 A1 WO2022062567 A1 WO 2022062567A1
Authority
WO
WIPO (PCT)
Prior art keywords
test data
test
target
current
preset
Prior art date
Application number
PCT/CN2021/104882
Other languages
English (en)
French (fr)
Inventor
徐琨
Original Assignee
爱德万测试股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 爱德万测试股份有限公司 filed Critical 爱德万测试股份有限公司
Publication of WO2022062567A1 publication Critical patent/WO2022062567A1/zh
Priority to US18/119,844 priority Critical patent/US20230214568A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/32—Circuit design at the digital level
    • G06F30/333—Design for testability [DFT], e.g. scan chain or built-in self-test [BIST]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming

Definitions

  • the invention relates to the technical field of chip testing, in particular to a detection method, system, electronic device and storage medium for product test data.
  • the parametric test results usually conform to a normal distribution. At present, it is mainly judged whether the test result is within the test specification range, and if so, the chip is determined to pass the test (that is, the chip quality is qualified). When these abnormal chips will still be shipped as good products, they will affect the quality of chip manufacturing and even cause quality accidents. Therefore, the existing detection methods of chip test data cannot meet higher chip production requirements.
  • the technical problem to be solved by the present invention is to overcome the defect that the method for mass production testing of semiconductor chips in the prior art cannot meet actual needs, and the purpose is to provide a detection method, system, electronic device and storage medium for product test data .
  • the present invention provides a detection method for product test data, the detection method includes:
  • the intermediate test data is grouped according to different preset test parameters to obtain a plurality of first groups; wherein each of the preset test parameters corresponds to one of the first groups;
  • the target test limit is used to test the test data of a new batch of products.
  • the step of obtaining the target test limit value according to the intermediate test data corresponding to the target grouping includes:
  • Statistical parameters corresponding to the target grouping are obtained by calculating according to the intermediate test data corresponding to the target grouping; wherein, the statistical parameters include an average value and a mean square error;
  • the upper test limit value and the lower test limit value are calculated according to the statistical parameters and preset constraints, and the upper test limit value and the lower test limit value are used as the target test limit value.
  • the step further includes:
  • the detection method further includes:
  • the initial population parameter it is determined whether the current test data falls within the center region of the normal distribution corresponding to the current training population, and if so, it is determined that the robustness of the current test data meets preset requirements, and the inserting the current test data into the training population to form a target training population;
  • the target population parameter it is determined whether the current test data falls within the center area of the normal distribution corresponding to the target training population, and if so, it is determined that the robustness of the current test data meets the preset requirements, and the The current test data is inserted into the target training population to form a new target training population;
  • the target test limit is updated according to the new test data corresponding to the target training population.
  • the detection method also includes:
  • the difference between the statistical parameters of the target training population before the update and the updated target training population is less than a first set threshold, and the test data corresponding to the updated target training population does not satisfy the preset Conditions; the statistical parameters include mean and mean square error.
  • the step of judging whether the test data in the target training population satisfies a preset condition, and if so, generating first test data to update the target training population includes:
  • the first test data is randomly generated to update the target training population.
  • the step of randomly generating the first test data to update the target training population includes:
  • At least one of the inverse function sampling method, the Box-Muller transformation method (a method for generating normally distributed random numbers), and the central limit theorem is used to randomly generate a set of second test data, respectively, and calculate the The difference between the statistical parameter corresponding to the second test data and the statistical parameter of the target training population before the update, and the second test data corresponding to the minimum absolute value of the difference is selected as the first Test data to update the target training population.
  • the step of screening the historical test data to obtain intermediate test data includes:
  • Test data exceeding a preset test limit is excluded from the third test data to obtain the intermediate test data.
  • the step of grouping the intermediate test data to obtain a plurality of first groups according to preset test parameters includes:
  • the step further includes:
  • the steps include:
  • the step of filtering out the third test data corresponding to all the preset test parameters in the historical test data includes:
  • the third test data is output from the static data space through a different API (application programming interface).
  • the present invention also provides a detection system for product test data, the detection system includes:
  • the historical data acquisition module is used to acquire historical test data corresponding to multiple historical batches of products
  • an intermediate data acquisition module used for screening the historical test data to acquire intermediate test data
  • a grouping acquisition module configured to perform grouping processing on the intermediate test data according to different preset test parameters to obtain a plurality of first groups; wherein each of the preset test parameters corresponds to one of the first groups;
  • a distribution type acquisition module configured to acquire a first distribution type corresponding to each of the first groupings according to the intermediate test data corresponding to the first grouping;
  • a first judgment module configured to judge whether the first distribution type is a preset distribution type, and if so, use the first group corresponding to the first distribution type as a target group;
  • test limit obtaining module configured to obtain a target test limit according to the intermediate test data corresponding to the target grouping
  • the target test limit is used to test the test data of a new batch of products.
  • the test limit acquisition module includes:
  • a parameter calculation unit configured to calculate and obtain statistical parameters corresponding to the target grouping according to the intermediate test data corresponding to the target grouping; wherein, the statistical parameters include an average value and a mean square error;
  • a test limit calculation unit used for calculating the upper limit value and the lower limit value of the test according to the statistical parameters and preset constraints, and using the upper limit value of the test and the lower limit of the test as the target test limit.
  • the detection system further includes:
  • the current data acquisition module is used to acquire the current test data corresponding to the current test group in the current batch of products
  • a target data acquisition module for acquiring multiple groups of target test data corresponding to different preset test parameters in the current test data
  • the second judgment module is used for judging whether the target test data is within the corresponding target test limit, if yes, then determine that the target test data is normal test data; if not, determine that the target test data is abnormal test data;
  • the detection system when it is determined that the current test data corresponding to the current test group in the current batch of products passes the test, and the preset distribution type is normal distribution, the detection system further includes:
  • a current population acquisition module configured to use the intermediate test data corresponding to the target grouping as the current training population
  • a population parameter calculation module used to calculate and obtain the initial population parameter corresponding to the current training population
  • a third judging module configured to judge whether the current test data falls within the center area of the normal distribution corresponding to the current training population according to the initial population parameter, and if so, determine the robustness of the current test data Meet preset requirements, and insert the current test data into the training population to form a target training population;
  • a test limit update module for updating the target test limit according to the test data corresponding to the target training population
  • the population parameter calculation module is further configured to obtain the target population parameter corresponding to the target training population
  • the third judgment module is further configured to judge whether the current test data falls within the center region of the normal distribution corresponding to the target training population according to the target population parameter, and if so, determine the robustness of the current test data meeting preset requirements, and inserting the current test data into the target training population to form a new target training population;
  • the test limit update module is further configured to update the target test limit according to the new test data corresponding to the target training population.
  • the detection system further includes:
  • a fourth judgment module configured to judge whether the test data in the target training population satisfies a preset condition, and if so, generate first test data to update the target training population
  • the difference between the statistical parameters of the target training population before the update and the updated target training population is less than a first set threshold, and the test data corresponding to the updated target training population does not satisfy the preset Conditions; the statistical parameters include mean and mean square error.
  • the fourth judgment module includes:
  • a quartile acquiring unit configured to acquire the quartile corresponding to the test data in the target training population
  • a first judging unit for judging whether the first quartile in the quartile is equal to the third quartile, and if so, calling the generating unit;
  • the generating unit is configured to randomly generate the first test data to update the target training population.
  • the generating unit adopts at least one of the inverse function sampling method, the Box-Muller transformation method, and the central limit theorem to randomly generate a set of second test data, respectively, and calculate the corresponding value of each set of the second test data.
  • the intermediate data acquisition module includes:
  • a screening unit configured to screen out the third test data corresponding to all the preset test parameters in the historical test data
  • a rejecting unit configured to reject test data exceeding a preset test limit from the third test data to obtain the intermediate test data.
  • the grouping acquisition module includes:
  • a grouping unit configured to perform grouping processing on the intermediate test data according to different preset test parameters, and obtain a plurality of intermediate groups
  • the second judging unit is configured to judge whether the size of the intermediate group is greater than or equal to a second set threshold, and if so, the intermediate group is used as the first group.
  • the detection system includes:
  • a data space establishment module is used to pre-establish a static data space
  • a storage module configured to obtain the historical test data in a set format, decode the historical test data, and store the decoded historical test data in the static data space;
  • the screening unit is configured to output the third test data from the static data space through different APIs based on all the preset test parameters.
  • the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, the processor implementing the above-mentioned method for detecting product test data when executing the computer program.
  • the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the steps of the above-mentioned method for detecting product test data.
  • the initial test limit is obtained by calculation, that is, the key limit is tightened before the new batch of products is tested to ensure the detection accuracy of the test data of the new batch of products, so as to improve the test quality of the chip; when the new batch of products is tested
  • the robustness of the test data satisfies the set conditions, the current test data is inserted into the previous training population to form a new training population, and then updated to obtain a new dynamic test limit, that is, to achieve adaptive control of the test limit.
  • the dynamic adjustment of the chip can effectively detect the chip test data with abnormal data in real time, thereby further improving the test quality of the chip.
  • FIG. 1 is a flowchart of a method for detecting product test data according to Embodiment 1 of the present invention.
  • FIG. 2 is a schematic diagram of a test limit in a method for detecting product test data according to Embodiment 1 of the present invention.
  • FIG. 3 is a first flow chart of a method for detecting product test data according to Embodiment 2 of the present invention.
  • FIG. 4 is a second flowchart of the method for detecting product test data according to Embodiment 2 of the present invention.
  • FIG. 5 is a schematic diagram of the normal distribution of the training population in the method for detecting product test data according to Embodiment 2 of the present invention.
  • FIG. 6 is a schematic diagram of a process of generating a new training population in the method for detecting product test data according to Embodiment 2 of the present invention.
  • FIG. 7 is a first test schematic diagram of a method for detecting product test data according to Embodiment 2 of the present invention.
  • FIG. 8 is a second test schematic diagram of the method for detecting product test data according to Embodiment 2 of the present invention.
  • FIG. 9 is a schematic diagram of a detection result corresponding to an existing dynamic DPAT detection method.
  • FIG. 10 is a schematic diagram of a detection result corresponding to the detection method of product test data according to Embodiment 2 of the present invention.
  • FIG. 11 is a schematic block diagram of a system for detecting product test data according to Embodiment 3 of the present invention.
  • FIG. 12 is a schematic block diagram of a system for detecting product test data according to Embodiment 4 of the present invention.
  • FIG. 13 is a schematic structural diagram of an electronic device implementing a method for detecting product test data according to Embodiment 5 of the present invention.
  • the detection method of the product test data of the present embodiment includes:
  • the historical test data is the test data of at least six historical batches of products that pass the test limit defined by the equipment specification, and each batch of products includes at least 30 test parameters; wherein, the historical test data corresponds to The number of historical batches of products and the number of detection parameters in each batch of products can be re-determined and adjusted according to the actual situation.
  • STDF Standard Test Data Format
  • a static data space needs to be established in the memory in advance, and once the static data space is established, the mass production test process can be started.
  • the STDF file corresponding to the historical test data is initialized, and the STDF content corresponding to the STDF file is decoded and stored in the static data space; wherein, the decoded STDF content is ASCII data.
  • the third test data corresponding to all preset test parameters in the historical test data is filtered out, and the test data exceeding the preset test limit is eliminated from the third test data to obtain the intermediate test data.
  • the third test data is output from the static data space through different APIs.
  • Preset test parameters include, but are not limited to, test items or multiple homology application sites.
  • the intermediate test data is grouped according to different preset test parameters, multiple intermediate groups are acquired, and it is judged whether the size of the intermediate group is greater than or equal to the second set threshold, and if so, the intermediate group is used as the The first grouping, that is, excluding the grouping containing a smaller amount of data, reduces the overall calculation amount, improves the calculation efficiency, and further improves the overall detection efficiency.
  • the preset distribution type includes normal distribution, that is, through the comparison of distribution types, the grouping of other distribution types such as 0-1 distribution is eliminated, and only the normal distribution grouping is retained, thereby ensuring that the later test limit is determined. Accuracy and reliability.
  • the target test limit is used for mass production testing of the test data of the new batch of products.
  • the target test limit is used as the first dynamic limit to ensure that the critical limit is tightened before starting the new batch of product testing, and the detection accuracy of the test data of the new batch of product is guaranteed.
  • the statistical parameters corresponding to the target grouping are calculated and obtained according to the intermediate test data corresponding to the target grouping; wherein, the statistical parameters include an average value and a mean square error;
  • CPK constraints that is, process capability index
  • the test limit value is [0, 200] based only on the design specification of the product to be tested; as shown in (b) of FIG.
  • Statistical parameters and CPK constraints ( ⁇ CPK*sigma, sigma means mean square error) to calculate the upper and lower test values [46.13, 67.87], that is, a tighter test can be calculated by combining the CPK constraints limit to ensure the test quality of the chip.
  • the detection method of the product test data of the present embodiment is a further improvement to the embodiment 1, specifically:
  • step S106 it also includes:
  • S1010 When the set number of target test data are all normal test data, determine that the current test data of the current test group passes the test; otherwise, determine that the current test data of the current test group fails the test.
  • the set number of target test data can be all target test data, or can be determined according to the actual situation. For example, when 98 of the 100 target test data are normal test data, the current test group of The test data passed the test.
  • the step S1010 further includes:
  • S1013 determine whether the current test data falls within the center region of the normal distribution corresponding to the current training population, and if so, determine that the robustness of the current test data meets the preset requirements, and insert the current test data into the current test data. training the population to form the target training population;
  • a1 represents the adaptability of the data
  • a2 represents normal fitting (normal distribution fitting curve)
  • a3 represents +3sigma
  • a4 represents -3sigma
  • MEAT-LL represents the lower limit of the test
  • MEAT-UL represents the test Upper limit.
  • the adaptive function is used to continuously monitor the test data of each chip, and the purpose of adaptive testing is achieved with the continuous development of the training population.
  • the corresponding test data in the current test data meets the robustness requirement, the corresponding test data in the current test data is inserted into the previous training population to form the target training population.
  • the robustness can be determined by the following formula:
  • the target test limit is updated in time through the test data corresponding to the new test group in the same batch of products to ensure the quality of the chip test.
  • test data of the current chip is continuously used as a new individual, and the fitness function is used to compare it with the population array to evaluate its robustness.
  • the detection method of this embodiment belongs to a real-time test data monitoring algorithm based on evolutionary theory, called MEAT (Monitoring Evolutionary Algorithm under Test), which combines the characteristics of static PAT (Part Average Test Guide) and dynamic PAT, and introduces CPK constraints and evolution strategies are developed to achieve high-quality testing of consumer chips.
  • MEAT Monitoring Evolutionary Algorithm under Test
  • the horizontal axis represents the test data sequence
  • the vertical axis Test Data Distribution represents the test data range
  • LL represents the lower limit of the test
  • UL represents the upper limit of the test
  • the dots in the A area represent each current test data; It is known that the current test data is detected based on the target test limit obtained above.
  • MEAT monitors each test item as a separate training population; when the preset test parameters include multiple sites in the homology application, MEAT monitors multiple homology applications. Each site in the source application is monitored as a separate training population. Among them, as shown in Figure 8, for the dynamic limit under multi-site application, each site has an independent limit line.
  • step S1017 it also includes:
  • the difference between the statistical parameters of the target training population before the update and the target training population after the update is less than the first set threshold, and the test data corresponding to the target training population after the update does not meet the preset conditions; the statistical parameters include the average value and mean square error.
  • Step S1018 specifically includes:
  • At least one of the inverse function sampling method, the Box-Muller transformation method, and the central limit theorem is used to randomly generate a set of second test data, respectively, and calculate the statistical parameters corresponding to each set of second test data and the pre-update data.
  • the difference between the statistical parameters of the target training population, and the second test data corresponding to the smallest absolute value of the difference is selected as the first test data to update the target training population.
  • a method capable of randomly generating data can also be used to generate test data.
  • Using randomly generated test data to replace the test data in the original target training population can effectively avoid the local convergence of the population during the evolution process so that the UL and LL are too close, thereby ensuring the reliability of the dynamic test limit. .
  • the detection method MEAT of this embodiment does not need to be based on other information other than the above content, such as the coordinates of the die on the wafer, thereby improving the detection efficiency and accuracy of the existing product detection method.
  • FIG. 9 it is the detection result of detecting the test data based on the existing Dynamic PAT (Dynamic PAT) detection method.
  • the horizontal axis represents the test data sequence
  • the vertical axis represents the test data range.
  • DPAT-LL indicates the lower limit of the test
  • DPAT-UL indicates the upper limit of the test.
  • the DPAT-UL also increased significantly. It can be seen that this detection method has a high dependence on the test data, so the data is continuously released. Then its detection mechanism will have a greater impact, and may even lose its effectiveness.
  • the MEAT detection method decides whether to perform population evolution according to the robustness of the data, which reduces the sensitivity of the MEAT dynamic limit to the test data, and has a more reasonable mechanism to reduce the dependence on the test data. Under this detection method Data can even be published continuously, effectively tightening dynamic limits effectively during production.
  • test data detection method of this embodiment can effectively The abnormal test data is analyzed, thereby effectively improving the test quality of the chip.
  • MEAT detection methods can even be applied to data consistency checks across operating systems and programming languages, enabling traceable adaptive data creation and real-time storage, etc.
  • the target is filtered out according to the distribution type of the test data in each group.
  • the detection system for product test data in this embodiment includes a historical data acquisition module 1, an intermediate data acquisition module 2, a group acquisition module 3, a distribution type acquisition module 4, a first judgment module 5 and a test limit acquisition module Module 6.
  • the historical data acquisition module 1 is used to acquire historical test data corresponding to multiple historical batches of products
  • the historical test data is the test data of at least six historical batches of products that pass the test limit defined by the equipment specification, and each batch of products includes at least 30 test parameters; wherein, the historical test data corresponds to The number of historical batches of products and the number of detection parameters in each batch of products can be re-determined and adjusted according to the actual situation.
  • STDF file which belongs to batch production test data file; of course, it can also be stored in other format files according to the actual situation.
  • a static data space needs to be established in the memory in advance, and once the static data space is established, the mass production test process can be started.
  • the STDF file corresponding to the historical test data is initialized, and the STDF content corresponding to the STDF file is decoded and stored in the static data space; wherein, the decoded STDF content is ASCII data.
  • the detection system for product test data in this embodiment further includes a data space establishment module and a storage module.
  • the data space establishment module is used for pre-establishing the static data space;
  • the storage module is used for acquiring the historical test data in the set format, decoding the historical test data, and storing the decoded historical test data in the static data space.
  • the intermediate data acquisition module 2 is used for screening the historical test data to acquire the intermediate test data
  • the intermediate data acquisition module 2 includes a screening unit and a rejecting unit.
  • the screening unit is used for screening out the third test data corresponding to all preset test parameters in the historical test data; the removing unit is used for removing the test data exceeding the preset test limit from the third test data to obtain the intermediate test data.
  • the third test data is output from the static data space through different APIs.
  • Preset test parameters include, but are not limited to, test items or multiple homology application sites.
  • the grouping acquisition module 3 is configured to perform grouping processing on the intermediate test data according to different preset test parameters to obtain a plurality of first groups; wherein, each preset test parameter corresponds to a first group;
  • the grouping acquisition module 3 includes a grouping unit and a second judging unit.
  • a grouping unit configured to perform grouping processing on the intermediate test data according to different preset test parameters, and obtain a plurality of intermediate groups
  • a second judging unit used to determine whether the size of the intermediate group is greater than or equal to the second set threshold, and if so, Then, the intermediate grouping is used as the first grouping, that is, the grouping containing a small amount of data is eliminated, the overall calculation amount is reduced, the calculation efficiency is improved, and the overall detection efficiency is improved.
  • the distribution type obtaining module 4 is used to obtain the first distribution type corresponding to each first grouping according to the intermediate test data corresponding to the first grouping;
  • the first judgment module 5 is used for judging whether the first distribution type is a preset distribution type, and if so, the first grouping corresponding to the first distribution type is used as the target grouping;
  • the preset distribution type includes normal distribution, that is, through the comparison of distribution types, the grouping of other distribution types such as 0-1 distribution is eliminated, and only the normal distribution grouping is retained, thereby ensuring that the later test limit is determined. Accuracy and reliability.
  • the test limit obtaining module 6 is used to obtain the target test limit according to the intermediate test data corresponding to the target grouping;
  • the target test limit is used to test the test data of the new batch of products.
  • the target test limit is used as the first dynamic limit to ensure that the critical limit is tightened before starting the new batch of product testing, and the detection accuracy of the test data of the new batch of product is guaranteed.
  • test limit acquisition module 6 includes a parameter calculation unit and a test limit calculation unit;
  • the parameter calculation unit is used for calculating the statistical parameters corresponding to the target grouping according to the intermediate test data corresponding to the target grouping; wherein, the statistical parameters include an average value and a mean square error;
  • the test limit calculation unit is used to calculate the test upper limit value and the test lower limit value according to the statistical parameters and preset constraints (CPK constraints, that is, the process capability index), and use the test upper limit value and the test lower limit value as the test upper limit value and the test lower limit value.
  • CPK constraints that is, the process capability index
  • the test limit value is [0, 200] based only on the design specification of the product to be tested; as shown in (b) of FIG.
  • Statistical parameters and CPK constraints ( ⁇ CPK*sigma, sigma means mean square error) to calculate the upper and lower test values [46.13, 67.87], that is, a tighter test can be calculated by combining the CPK constraints limit to ensure the test quality of the chip.
  • the detection system of the product test data of the present embodiment is a further improvement to the embodiment 3, specifically:
  • the detection system further includes a current data acquisition module 7 , a target data acquisition module 8 , a second judgment module 9 and a determination module 10 .
  • the current data acquisition module 7 is used to acquire the current test data corresponding to the current test group in the current batch of products
  • the target data acquisition module 8 is used for acquiring multiple groups of target test data corresponding to different preset test parameters in the current test data;
  • the second judging module 9 is used for judging whether the target test data is within the corresponding target test limit, and if so, then determine that the target test data is normal test data; if not, then determine that the target test data is abnormal test data;
  • the determination module 10 is configured to determine that the current test data of the current test group passes the test when the set number of target test data are normal test data; otherwise, determine that the current test data of the current test group fails the test.
  • the set number of target test data can be the entire number of target test data, or can be determined according to the actual situation, for example: when 98 of the 100 target test data are normal test data, the current test group The test data passed the test.
  • the detection system of this embodiment further includes a current population acquisition module 11, a population parameter calculation module 12, a Three judgment module 13 and test limit update module 14 .
  • the current population acquisition module 11 is configured to use the intermediate test data corresponding to the target grouping as the current training population;
  • the population parameter calculation module 12 is used to calculate the initial population parameter corresponding to the current training population
  • the third judging module 13 is configured to judge whether the current test data falls within the center region of the normal distribution corresponding to the current training population according to the initial population parameters, and if so, determine that the robustness of the current test data meets the preset requirements, and assign the current test data to the current training population. Insert test data into the training population to form the target training population;
  • the test limit update module 14 is used to update the target test limit according to the test data corresponding to the target training population
  • the adaptive function is used to continuously monitor the test data of each chip, and the purpose of adaptive testing is achieved with the continuous development of the training population.
  • the population parameter calculation module 12 is further configured to obtain the target population parameter corresponding to the target training population
  • the third judging module 13 is further configured to judge whether the current test data falls within the center area of the normal distribution corresponding to the target training population according to the target population parameters, and if so, determine the robustness of the current test data. meet the preset requirements, and insert the current test data into the target training population to form a new target training population;
  • the corresponding test data in the current test data meets the robustness requirement, the corresponding test data in the current test data is inserted into the previous training population to form the target training population.
  • the robustness can be determined by the following formula:
  • the test limit update module 14 is further configured to update the target test limit according to the new test data corresponding to the target training population.
  • the corresponding statistical parameters are calculated according to the test data corresponding to the new target training population, and the new target test limit is finally calculated in combination with the CPK constraints.
  • the target test limit is updated in time through the test data corresponding to the new test group in the same batch of products to ensure the quality of the chip test.
  • test data of the current chip is continuously used as a new individual, and the fitness function is used to compare it with the population array to evaluate its robustness.
  • the detection method of this embodiment belongs to a real-time test data monitoring algorithm based on evolutionary theory, which is called MEAT.
  • the algorithm combines the characteristics of static PAT and dynamic PAT, and introduces CPK constraints and evolution strategies, so as to realize the detection of consumer chips. high-quality testing.
  • the horizontal axis represents the test data sequence
  • the vertical axis Test Data Distribution represents the test data range
  • LL represents the lower limit of the test
  • UL represents the upper limit of the test
  • the dots in the A area represent each current test data; It is known that the current test data is detected based on the target test limit obtained above.
  • MEAT monitors each test item as a separate training population; when the preset test parameters include multiple sites in the homology application, MEAT monitors multiple homology applications. Each site in the source application is monitored as a separate training population. Among them, as shown in Figure 8, for dynamic limits under multi-site applications, each site has an independent limit line.
  • the detection system of this embodiment further includes a fourth judgment module 15;
  • the fourth judgment module 15 is used for judging whether the test data in the target training population satisfies the preset condition, and if so, generating first test data to update the target training population;
  • the difference between the statistical parameters of the target training population before the update and the target training population after the update is less than the first set threshold, and the test data corresponding to the target training population after the update does not meet the preset conditions; the statistical parameters include the average value and mean square error.
  • the fourth judgment module 15 includes a quartile acquisition unit, a first judgment unit and a generation unit.
  • the quartile obtaining unit is used to obtain the quartile corresponding to the test data in the target training population
  • the first judging unit is used to judge whether the first quartile in the quartile is equal to the third quartile, and if so, call the generating unit;
  • the generating unit is used for randomly generating the first test data to update the target training population.
  • the generating unit adopts at least one of the inverse function sampling method, the Box-Muller transformation method, and the central limit theorem to randomly generate a set of second test data, respectively, and calculate the statistical parameters and update corresponding to each set of second test data.
  • the difference between the statistical parameters of the previous target training population is selected, and the second test data corresponding to the smallest absolute value of the difference is selected as the first test data to update the target training population.
  • a method capable of randomly generating data can also be used to generate test data.
  • Using randomly generated test data to replace the test data in the original target training population can effectively avoid the local convergence of the population during the evolution process so that the UL and LL are too close, thereby ensuring the reliability of the dynamic test limit. .
  • the detection method MEAT of this embodiment does not need to be based on other information other than the above content, such as the coordinates of the die on the wafer, thereby improving the detection efficiency and accuracy of the existing product detection method.
  • FIG. 9 it is the detection result of detecting the test data based on the existing dynamic DPAT method.
  • the horizontal axis represents the test data sequence
  • the vertical axis represents the test data range.
  • DPAT-LL indicates the lower limit of the test
  • DPAT-UL indicates the upper limit of the test.
  • this detection method has a high dependence on the test data, so the data is continuously released. Then its detection mechanism will have a greater impact, and may even lose its effectiveness.
  • the MEAT detection method decides whether to perform population evolution according to the robustness of the data, which reduces the sensitivity of the MEAT dynamic limit to the test data, and has a more reasonable mechanism to reduce the dependence on the test data. Under this detection method Data can even be published continuously, effectively tightening dynamic limits effectively during production.
  • test data detection method of this embodiment can effectively The abnormal test data is analyzed, thereby effectively improving the test quality of the chip.
  • the target is filtered out according to the distribution type of the test data in each group.
  • FIG. 13 is a schematic structural diagram of an electronic device according to Embodiment 5 of the present invention.
  • the electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, the method for detecting product test data in any one of Embodiments 1 or 2 is implemented.
  • the electronic device 30 shown in FIG. 13 is only an example, and should not impose any limitation on the function and scope of use of the embodiment of the present invention.
  • the electronic device 30 may take the form of a general-purpose computing device, for example, it may be a server device.
  • Components of the electronic device 30 may include, but are not limited to, the above-mentioned at least one processor 31 , the above-mentioned at least one memory 32 , and a bus 33 connecting different system components (including the memory 32 and the processor 31 ).
  • the bus 33 includes a data bus, an address bus and a control bus.
  • Memory 32 may include volatile memory, such as random access memory (RAM) 321 and/or cache memory 322, and may further include read only memory (ROM) 323.
  • RAM random access memory
  • ROM read only memory
  • the memory 32 may also include a program/utility 325 having a set (at least one) of program modules 324 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, which An implementation of a network environment may be included in each or some combination of the examples.
  • the processor 31 executes various functional applications and data processing by running the computer program stored in the memory 32, such as the detection method of product test data in any one of Embodiments 1 or 2 of the present invention.
  • the electronic device 30 may also communicate with one or more external devices 34 (eg, keyboards, pointing devices, etc.). Such communication may take place through input/output (I/O) interface 35 .
  • the model-generating device 30 may also communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN), and/or a public network such as the Internet) through a network adapter 36 . As shown in FIG. 13 , the network adapter 36 communicates with other modules of the model generation device 30 via the bus 33 .
  • networks eg, a local area network (LAN), a wide area network (WAN), and/or a public network such as the Internet
  • model-generated device 30 may be used in conjunction with the model-generated device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk) array) systems, tape drives, and data backup storage systems.
  • This embodiment provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the steps in the method for detecting product test data in any one of Embodiments 1 or 2.
  • the readable storage media may include, but are not limited to, portable disks, hard disks, random access memories, read-only memories, erasable programmable read-only memories, optical storage devices, magnetic storage devices, or any of the above suitable combination.
  • the present invention can also be implemented in the form of a program product, which includes program codes.
  • the program product runs on a terminal device, the program code is used to make the terminal device execute the implementation of the implementation in Embodiment 1 or 2. Steps in the method for detecting product test data in any one of the embodiments.
  • the program code for executing the present invention can be written in any combination of one or more programming languages, and the program code can be completely executed on the user equipment, partially executed on the user equipment, as an independent software
  • the package executes, partly on the user device, partly on the remote device, or entirely on the remote device.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Optimization (AREA)
  • Software Systems (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Geometry (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Genetics & Genomics (AREA)
  • Artificial Intelligence (AREA)
  • Algebra (AREA)
  • Computational Linguistics (AREA)
  • Operations Research (AREA)
  • Databases & Information Systems (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Physiology (AREA)
  • Complex Calculations (AREA)
  • Testing Or Measuring Of Semiconductors Or The Like (AREA)
  • General Factory Administration (AREA)

Abstract

一种产品测试数据的检测方法、系统、电子设备和存储介质，该检测方法包括获取多个历史批次产品的历史测试数据；对历史测试数据进行筛选以获取中间测试数据；根据预设测试参数对中间测试数据进行分组以获取第一分组；根据第一分组的中间测试数据获取第一分组的分布类型；在分布类型为预设分布类型时将分布类型对应的第一分组作为目标分组；根据目标分组对应的中间测试数据获取目标测试限值。该方法基于历史测试数据确定初始测试限值，实现在新批次产品检测之前收紧关键限值，以保证对新批次产品的测试数据的检测准确性；能够自适应地对测试限值的动态调整，实时有效地检测出存在异常数据的芯片测试数据，提高了芯片的测试质量。

Description

产品测试数据的检测方法、系统、电子设备和存储介质
本申请要求申请日为2020年9月23日的中国专利申请202011007511.8的优先权。本申请引用上述中国专利申请的全文。
技术领域
本发明涉及芯片测试技术领域，特别涉及一种产品测试数据的检测方法、系统、电子设备和存储介质。
背景技术
在半导体芯片量产测试阶段，参数测试结果通常符合正态分布。目前，主要通过判断测试结果是否在测试规格范围内，若在则确定芯片通过测试(即芯片质量合格)；然而，有些芯片对应的测试结果即便在测试规格范围，但是会过于偏离平均值，当这些异常的芯片仍将作为良好产品装运时，就会影响芯片制造的质量，甚至造成质量事故，因此现有的芯片测试数据的检测方法无法满足更高的芯片生产需求。
发明内容
本发明要解决的技术问题是为了克服现有技术中对半导体芯片量产测试的方式存在不能够满足实际需求的缺陷，目的在于提供一种产品测试数据的检测方法、系统、电子设备和存储介质。
本发明是通过下述技术方案来解决上述技术问题：
本发明提供一种产品测试数据的检测方法，所述检测方法包括：
获取多个历史批次产品对应的历史测试数据；
对所述历史测试数据进行筛选处理以获取中间测试数据；
根据不同的预设测试参数对所述中间测试数据进行分组处理以获取多个第一分组；其中，每种所述预设测试参数对应一个所述第一分组；
根据所述第一分组对应的所述中间测试数据获取每个所述第一分组对应的第一分布类型；
判断所述第一分布类型是否为预设分布类型，若是，则将所述第一分布类型对应的所述第一分组作为目标分组；
根据所述目标分组对应的所述中间测试数据获取目标测试限值；
其中，所述目标测试限值用于对新批次产品的测试数据进行测试。
较佳地，所述根据所述目标分组对应的所述中间测试数据获取目标测试限值的步骤 包括：
根据所述目标分组对应的所述中间测试数据计算得到所述目标分组对应的统计学参数；其中，所述统计学参数包括平均值和均方差；
根据所述统计学参数和预设约束条件计算得到测试上限值和测试下限值，并将所述测试上限值和所述测试下限值作为所述目标测试限值。
较佳地，所述根据所述目标分组对应的所述中间测试数据获取目标测试限值的步骤之后还包括：
获取当前批次产品中当前测试组对应的当前测试数据；
获取所述当前测试数据中与不同的所述预设测试参数对应的多组目标测试数据；
判断所述目标测试数据是否在对应的所述目标测试限值内，若是，则确定所述目标测试数据为正常测试数据；若否，则确定所述目标测试数据为异常测试数据；
在设定数量的所述目标测试数据均为正常测试数据时，则确定当前测试组的所述当前测试数据通过检测；否则，确定当前测试组的所述当前测试数据未通过检测。
较佳地，在确定当前批次产品中当前测试组对应的所述当前测试数据通过检测，且所述预设分布类型为正态分布时，所述检测方法还包括：
将所述目标分组对应的所述中间测试数据作为当前训练种群；
计算得到所述当前训练种群对应的初始种群参数；
根据所述初始种群参数判断所述当前测试数据是否落入所述当前训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所述当前测试数据插入至所述训练种群以形成目标训练种群；
根据所述目标训练种群对应的测试数据更新所述目标测试限值；
对于当前批次产品中下一测试组对应的测试数据，计算得到所述目标训练种群对应的目标种群参数；
根据所述目标种群参数判断当前测试数据是否落入所述目标训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所述当前测试数据插入至所述目标训练种群以形成新的所述目标训练种群；
根据新的所述目标训练种群对应的测试数据更新所述目标测试限值。
较佳地，所述检测方法还包括：
判断所述目标训练种群中的测试数据是否满足预设条件，若满足，则生成第一测试数据以更新所述目标训练种群；
其中，更新前的所述目标训练种群与更新后的所述目标训练种群的统计学参数相差小于第一设定阈值，且更新后的所述目标训练种群对应的测试数据不满足所述预设条件；所述统计学参数包括平均值和均方差。
较佳地，所述判断所述目标训练种群中的测试数据是否满足预设条件，若满足，则生成第一测试数据以更新所述目标训练种群的步骤包括：
获取所述目标训练种群中的测试数据对应的四分位数；
判断所述四分位数中的第一四分位数是否等于第三四分位数，若是，则随机生成所述第一测试数据以更新所述目标训练种群。
较佳地，所述随机生成所述第一测试数据以更新所述目标训练种群的步骤包括：
采用反函数采样方法、Box-Muller变换方法(一种生成正态分布的随机数的方法)、中央极限定理中的至少一种方式，分别随机生成一组第二测试数据，计算每组所述第二测试数据对应的统计学参数与更新前的所述目标训练种群的统计学参数的差值，并选取所述差值的绝对值最小时对应的所述第二测试数据作为所述第一测试数据以更新所述目标训练种群。
较佳地，所述对所述历史测试数据进行筛选处理以获取中间测试数据的步骤包括：
筛选出所述历史测试数据中与所有所述预设测试参数对应的第三测试数据；
从所述第三测试数据中剔除超出预设测试限值的测试数据以获取所述中间测试数据。
较佳地，所述根据预设测试参数对所述中间测试数据进行分组处理以获取多个第一分组的步骤包括：
根据不同的所述预设测试参数对所述中间测试数据进行分组处理，获取多个中间分组；
判断所述中间分组的大小是否大于或者等于第二设定阈值，若是，则将所述中间分组作为所述第一分组。
较佳地，所述获取多个历史批次产品对应的历史测试数据的步骤之前还包括：
预先建立静态数据空间；
所述获取多个历史批次产品对应的历史测试数据的步骤之后、所述对所述历史测试数据进行筛选处理以获取中间测试数据的步骤之前包括：
获取设定格式的所述历史测试数据，并对所述历史测试数据进行解码处理并将解码后的所述历史测试数据存储至所述静态数据空间；
所述筛选出所述历史测试数据中与所有所述预设测试参数对应的第三测试数据的步骤包括：
基于所有所述预设测试参数，通过不同的API(应用程序接口)从所述静态数据空间输出所述第三测试数据。
本发明还提供一种产品测试数据的检测系统，所述检测系统包括：
历史数据获取模块，用于获取多个历史批次产品对应的历史测试数据；
中间数据获取模块，用于对所述历史测试数据进行筛选处理以获取中间测试数据；
分组获取模块，用于根据不同的预设测试参数对所述中间测试数据进行分组处理以获取多个第一分组；其中，每种所述预设测试参数对应一个所述第一分组；
分布类型获取模块，用于根据所述第一分组对应的所述中间测试数据获取每个所述第一分组对应的第一分布类型；
第一判断模块，用于判断所述第一分布类型是否为预设分布类型，若是，则将所述第一分布类型对应的所述第一分组作为目标分组；
测试限值获取模块，用于根据所述目标分组对应的所述中间测试数据获取目标测试限值；
其中，所述目标测试限值用于对新批次产品的测试数据进行测试。
较佳地，所述测试限值获取模块包括：
参数计算单元，用于根据所述目标分组对应的所述中间测试数据计算得到所述目标分组对应的统计学参数；其中，所述统计学参数包括平均值和均方差；
测试限值计算单元，用于根据所述统计学参数和预设约束条件计算得到测试上限值和测试下限值，并将所述测试上限值和所述测试下限值作为所述目标测试限值。
较佳地，所述检测系统还包括：
当前数据获取模块，用于获取当前批次产品中当前测试组对应的当前测试数据；
目标数据获取模块，用于获取所述当前测试数据中与不同的所述预设测试参数对应的多组目标测试数据；
第二判断模块，用于判断所述目标测试数据是否在对应的所述目标测试限值内，若是，则确定所述目标测试数据为正常测试数据；若否，则确定所述目标测试数据为异常测试数据；
确定模块，用于在设定数量的所述目标测试数据均为正常测试数据时，则确定当前测试组的所述当前测试数据通过检测；否则，确定当前测试组的所述当前测试数据未通过检测。
较佳地，在确定当前批次产品中当前测试组对应的所述当前测试数据通过检测，且所述预设分布类型为正态分布时，所述检测系统还包括：
当前种群获取模块，用于将所述目标分组对应的所述中间测试数据作为当前训练种群；
种群参数计算模块，用于计算得到所述当前训练种群对应的初始种群参数；
第三判断模块，用于根据所述初始种群参数判断所述当前测试数据是否落入所述当前训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所述当前测试数据插入至所述训练种群以形成目标训练种群；
测试限值更新模块，用于根据所述目标训练种群对应的测试数据更新所述目标测试 限值；
对于当前批次产品中下一测试组对应的测试数据，所述种群参数计算模块还用于计算得到所述目标训练种群对应的目标种群参数；
所述第三判断模块还用于根据所述目标种群参数判断当前测试数据是否落入所述目标训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所述当前测试数据插入至所述目标训练种群以形成新的所述目标训练种群；
所述测试限值更新模块还用于根据新的所述目标训练种群对应的测试数据更新所述目标测试限值。
较佳地，所述检测系统还包括：
第四判断模块，用于判断所述目标训练种群中的测试数据是否满足预设条件，若满足，则生成第一测试数据以更新所述目标训练种群；
其中，更新前的所述目标训练种群与更新后的所述目标训练种群的统计学参数相差小于第一设定阈值，且更新后的所述目标训练种群对应的测试数据不满足所述预设条件；所述统计学参数包括平均值和均方差。
较佳地，所述第四判断模块包括：
四分位数获取单元，用于获取所述目标训练种群中的测试数据对应的四分位数；
第一判断单元，用于判断所述四分位数中的第一四分位数是否等于第三四分位数，若是，则调用生成单元；
所述生成单元用于随机生成所述第一测试数据以更新所述目标训练种群。
较佳地，所述生成单元采用反函数采样方法、Box-Muller变换方法、中央极限定理中的至少一种方式，分别随机生成一组第二测试数据，计算每组所述第二测试数据对应的统计学参数与更新前的所述目标训练种群的统计学参数的差值，并选取所述差值的绝对值最小时对应的所述第二测试数据作为所述第一测试数据以更新所述目标训练种群。
较佳地，所述中间数据获取模块包括：
筛选单元，用于筛选出所述历史测试数据中与所有所述预设测试参数对应的第三测试数据；
剔除单元，用于从所述第三测试数据中剔除超出预设测试限值的测试数据以获取所述中间测试数据。
较佳地，所述分组获取模块包括：
分组单元，用于根据不同的所述预设测试参数对所述中间测试数据进行分组处理，获取多个中间分组；
第二判断单元，用于判断所述中间分组的大小是否大于或者等于第二设定阈值，若 是，则将所述中间分组作为所述第一分组。
较佳地，所述检测系统包括：
数据空间建立模块，用于预先建立静态数据空间；
存储模块，用于获取设定格式的所述历史测试数据，并对所述历史测试数据进行解码处理并将解码后的所述历史测试数据存储至所述静态数据空间；
所述筛选单元用于基于所有所述预设测试参数，通过不同的API从所述静态数据空间输出所述第三测试数据。
本发明还提供一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行计算机程序时实现上述的产品测试数据的检测方法。
本发明还提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现上述的产品测试数据的检测方法的步骤。
在符合本领域常识的基础上，上述各优选条件，可任意组合，即得本发明各较佳实例。
本发明的积极进步效果在于：
本发明中，基于收集的历史若干批次产品的历史测试数据以及预设测试参数(如测试项)，对其进行过滤、分组等处理，根据每个分组中测试数据的分布类型筛选出目标分组以计算得到初始测试限值，即实现在新批次产品检测之前收紧关键限值，以保证对新批次产品的测试数据的检测准确性，以提高芯片的测试质量；当新批次产品的测试数据的鲁棒性满足设定条件时则将当前测试数据插入至先前的训练种群中以形成新的训练种群，进而更新得到新的动态测试限值，即实现自适应地对测试限值的动态调整，能够实时有效地检测出存在异常数据的芯片测试数据，从而进一步地提高了芯片的测试质量。
附图说明
图1为本发明实施例1的产品测试数据的检测方法的流程图。
图2为本发明实施例1的产品测试数据的检测方法中测试限值的示意图。
图3为本发明实施例2的产品测试数据的检测方法的第一流程图。
图4为本发明实施例2的产品测试数据的检测方法的第二流程图。
图5为本发明实施例2的产品测试数据的检测方法中训练种群的正态分布示意图。
图6为本发明实施例2的产品测试数据的检测方法中生成新的训练种群的过程示意图。
图7为本发明实施例2的产品测试数据的检测方法的第一测试示意图。
图8为本发明实施例2的产品测试数据的检测方法的第二测试示意图。
图9为现有的动态DPAT的检测方式对应的检测结果示意图。
图10为本发明实施例2的产品测试数据的检测方法对应的检测结果示意图。
图11为本发明实施例3的产品测试数据的检测系统的模块示意图。
图12为本发明实施例4的产品测试数据的检测系统的模块示意图。
图13为本发明实施例5的实现产品测试数据的检测方法的电子设备的结构示意图。
具体实施方式
下面通过实施例的方式进一步说明本发明，但并不因此将本发明限制在所述的实施例范围之中。
实施例1
如图1所示，本实施例的产品测试数据的检测方法包括：
S101、获取多个历史批次产品对应的历史测试数据；
在一可实施的方式中，历史测试数据为通过设备规范定义的测试限值的至少六个历史批次产品的测试数据，每个批次产品包括至少30个检测参数；其中，历史测试数据对应的历史批次产品的数量以及每个批次产品中检测参数的数量均可以根据实际情况进行重新确定与调整。
一般将历史测试数据存储为STDF(标准测试数据格式)文件，属于批量生产测试数据文件；当然也可以根据实际情况将其存储为其他格式文件。
另外，需要预先在内存中建立静态数据空间，一旦建立静态数据空间就可以开始批量生产测试流程。
在开始检测时，初始化历史测试数据对应的STDF文件，解码出STDF文件对应的STDF内容并存储至静态数据空间；其中，解码得到的STDF内容为ASCII数据。
S102、对历史测试数据进行筛选处理以获取中间测试数据；
其中，筛选出历史测试数据中与所有预设测试参数对应的第三测试数据，并从第三测试数据中剔除超出预设测试限值的测试数据以获取中间测试数据。
具体地，基于所有预设测试参数，通过不同的API从静态数据空间输出第三测试数据。预设测试参数包括但不限于测试项或多个同源性应用程序中站点。
另外，需要通过人工或自动方式检查获取的所有中间测试数据分布情况是否合理，排除明显不合理的测试数据，以保证后期测试限值确定的准确性和可靠性。
S103、根据不同的预设测试参数对中间测试数据进行分组处理以获取多个第一分组；其中，每种预设测试参数对应一个第一分组；
具体地，根据不同的预设测试参数对中间测试数据进行分组处理，获取多个中间分 组，并判断中间分组的大小是否大于或者等于第二设定阈值，若是，则将所述中间分组作为所述第一分组，即剔除包含较少数据量的分组，减少了整体计算量，提高了计算效率，进而提高了整体的检测效率。
S104、根据第一分组对应的中间测试数据获取每个第一分组对应的第一分布类型；
S105、判断第一分布类型是否为预设分布类型，若是，则将第一分布类型对应的第一分组作为目标分组；
其中，预设分布类型包括正态分布，即通过对分布类型的比对，将0-1分布等其他分布类型的分组剔除，仅保留正态分布的分组，从而保证了后期测试限值确定的准确性和可靠性。
S106、根据目标分组对应的中间测试数据获取目标测试限值；
其中，目标测试限值用于对新批次产品的测试数据进行量产测试。目标测试限值作为第一个动态限值，保证了开始新批次产品检测之前收紧关键限值，保证了对新批次产品的测试数据的检测准确性。
具体地，根据目标分组对应的中间测试数据计算得到目标分组对应的统计学参数；其中，统计学参数包括平均值和均方差；
根据统计学参数和预设约束条件(CPK约束条件，即过程能力指数)计算得到测试上限值和测试下限值，并将测试上限值和测试下限值作为目标测试限值。
例如：
Figure PCTCN2021104882-appb-000001
Figure PCTCN2021104882-appb-000002
其中，
Figure PCTCN2021104882-appb-000003
表示平均值，σ表示均方差，CPK为预设约束条件，Dynamic UL表示测试上限值，Dynamic LL表示测试下限值。
在一可实施的方式中，如图2中的(a)所示，仅基于待测产品设计规范得到测试限值为[0，200]；如图2中的(b)所示，当基于统计学参数和CPK约束条件(±CPK*sigma，sigma表示均方差)以计算得到测试上限值和测试下限值[46.13，67.87]，即通过结合CPK约束条件能够计算得到更收紧的测试限值，以保证芯片的测试质量。
本实施例中，基于收集的历史若干批次产品的历史测试数据以及预设测试参数(如测试项)，对其进行过滤、分组等处理，根据每个分组中测试数据的分布类型筛选出目标分组以计算得到测试限值，实现在新批次产品检测之前收紧关键限值，以保证对新批次产品的测试数据的检测准确性，从而提高芯片的测试质量。
实施例2
如图3所示，本实施例的产品测试数据的检测方法是对实施例1的进一步改进，具体地：
步骤S106之后还包括：
S107、获取当前批次产品中当前测试组对应的当前测试数据；
S108、获取当前测试数据中与不同的预设测试参数对应的多组目标测试数据；
S109、判断目标测试数据是否在对应的目标测试限值内，若是，则确定目标测试数据为正常测试数据；若否，则确定目标测试数据为异常测试数据；
S1010、在设定数量的目标测试数据均为正常测试数据时，则确定当前测试组的当前测试数据通过检测；否则，确定当前测试组的当前测试数据未通过检测。
其中，设定数量的目标测试数据可以为全部数量的目标测试数据，也可以根据实际情况具体确定，例如：当100个目标测试数据中有98个为正常测试数据时则确定当前测试组的当前测试数据通过检测。
如图4所示，在确定当前批次产品中当前测试组对应的当前测试数据通过检测，且预设分布类型为正态分布时，步骤S1010之后还包括：
S1011、将目标分组对应的中间测试数据作为当前训练种群；
S1012、计算得到当前训练种群对应的初始种群参数；
S1013、根据初始种群参数判断当前测试数据是否落入当前训练种群对应的正态分布的中心区域内，若是，则确定当前测试数据的鲁棒性满足预设要求，并将当前测试数据插入至当前训练种群以形成目标训练种群；
S1014、根据目标训练种群对应的测试数据更新目标测试限值；
如图5所示，a1表示数据的适应能力，a2表示normal fitting(正态分布拟合曲线)，a3表示+3sigma，a4表示-3sigma，MEAT-LL表示测试下限值，MEAT-UL表示测试上限值。
在当前测试数据落入当前训练种群对应的正态分布的中心区域时，则说明当前测试数据的鲁棒性(适应性)足够强；如图6所示，此时将其插入之前的训练种群中以形成新的训练种群，对应新的统计学参数，从而实现动态建立新的目标测试限值。
其中，在产片批量测试阶段，适应性功能用于连续监视每个芯片的测试数据，随着训练种群不断发展以达到自适应测试的目的。
S1015、对于当前批次产品中下一测试组对应的测试数据，计算得到所述目标训练种群对应的目标种群参数；
S1016、根据所述目标种群参数判断当前测试数据是否落入所述目标训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所述当前测试数据插入至所述目标训练种群以形成新的所述目标训练种群；
其中，对于同一预设测试参数，在当前测试数据中对应的测试数据满足鲁棒性要求时，则将当前测试数据中对应的测试数据插入之前的训练种群中以形成目标训练种群。
具体地，鲁棒性可以通过如下公式确定：
Figure PCTCN2021104882-appb-000004
S1017、根据新的所述目标训练种群对应的测试数据更新所述目标测试限值。具体地，根据新的目标训练种群对应的测试数据计算得到对应的统计学参数，并结合CPK约束条件最终计算得到新的目标测试限值。
通过同一批次产品中的新的测试组对应的测试数据及时更新目标测试限值以保证对芯片测试的质量。
即本实施例在量产自动测试过程中，源源不断地将当前芯片的测试数据将作为新个体，通过适应度函数与种群数组进行对比，以评估其鲁棒性。
本实施例的检测方法属于基于进化理论的实时测试数据监控演算法，称为MEAT(测试中监视进化演算法)，该算法结合了静态PAT(零件平均测试指南)和动态PAT的特性，并引入了CPK约束条件和演化策略，以实现对消费性芯片的高质量测试。
如图7所示，横轴表示测试数据序列，纵轴Test Data Distribution表示测试数据范围，LL表示测试下限值，UL表示测试上限值，A区域的圆点表示每个当前测试数据；可以得知，基于上述得到的目标测试限值对当前测试数据进行检测。
另外，当预设测试参数包括测试项时，MEAT将每个测试项目作为一个单独的训练种群进行监视；当预设测试参数包括多个同源性应用程序中站点时，则MEAT将多个同源性应用程序中每个站点作为一个单独的训练种群进行监视。其中，如图8所示，对于多站点应用下的动态限值，每个站点都有独立的限值线。
另外，步骤S1017之后还包括：
S1018、判断目标训练种群中的测试数据是否满足预设条件，若满足，则生成第一测试数据以更新目标训练种群；
其中，更新前的目标训练种群与更新后的目标训练种群的统计学参数相差小于第一设定阈值，且更新后的目标训练种群对应的测试数据不满足预设条件；统计学参数包括平均值和均方差。
步骤S1018具体包括：
获取目标训练种群中的测试数据对应的四分位数；
判断四分位数中的第一四分位数是否等于第三四分位数，若是，则随机生成第一测试数据以更新目标训练种群。
具体地，采用反函数采样方法、Box-Muller变换方法、中央极限定理中的至少一种 方式，分别随机生成一组第二测试数据，计算每组第二测试数据对应的统计学参数与更新前的目标训练种群的统计学参数的差值，并选取差值的绝对值最小时对应的第二测试数据作为第一测试数据以更新目标训练种群。当然也可以采用能够随机生成数据的方法来生成测试数据。
采用随机生成的测试数据替换原有的目标训练种群中的测试数据，可以有效地避免种群在进化过程中产生局部收敛以使UL和LL过于接近的情况发生，进而保证动态测试限值的可靠性。
本实施例的检测方法MEAT不需要基于上述内容之外的其他信息，例如晶粒在晶片上的坐标，从而提高了现有的产品检测方式的检测效率及准确度。
下面结合实例具体说明：
如图9所示，为基于现有Dynamic PAT(动态PAT)的检测方式对测试数据进行检测的检测结果。其中，横轴表示测试数据序列，纵轴表示测试数据范围。DPAT-LL表示测试下限值，DPAT-UL表示测试上限值。在图中的b1处，由于测试数据中出现连续异常数据，DPAT-UL也随之出现明显的抬升，由此可以得知，该检测方式对测试数据具有很高的依赖性，因此连续发布数据则将其检测机制产生较大的影响，甚至可能会失去效力。
如图10所示，为基于MEAT检测方法对测试数据进行检测的检测结果。其中，横轴表示测试数据序列，纵轴表示测试数据范围。MEAT-LL表示测试下限值，MEAT-UL表示测试上限值。可以得知，MEAT检测方法是根据数据稳健度决定是否进行种群进化，降低了MEAT动态限值对测试数据的敏感度，具有更合理的机制来减少对测试数据的依赖性，在该检测方式下甚至可以连续发布数据，在生产过程中能够有效地严格收紧动态限值。
动态DPAT的检测方式和本实施例的MEAT检测方法的检测方式的检测结果对比数据如下表：
待测芯片数量 DPPM 失效芯片数量 DPAT离群数 MEAT离群数
11440 50874 582 4 515
由上表可以得知，待测芯片数量为11440，失效芯片数量为582，则该批次芯片的DPPM(百万分比的缺陷率)＝(582/11440)*1000000＝50874。
采用现有DPAT的检测方式只能从发布的582个数据中检测出4个异常测试数据，对应的检测率＝(4/582)*100％＝0.69％，而采用本实施例的MEAT离群数能从发布的582个数据中检测出515个异常测试数据，对应的检测率＝(515/582)*100％＝88.49％，由此可知，采用本实施例的测试数据检测方法能够有效地分析出异常测试数据，从而有效地提高了芯片的测试质量。
另外，实验证明，本实施例中的MEAT检测方法对单元测试覆盖率超过C++和Java版本的95％。甚至还可以将MEAT检测方法应用在跨操作系统和编程语言的数据一致性检查，以实现可追溯适应性数据的创建和实时存储等。
本实施例中，基于收集的历史若干批次产品的历史测试数据以及预设测试参数(如测试项)，对其进行过滤、分组等处理，根据每个分组中测试数据的分布类型筛选出目标分组以计算得到初始测试限值；当新批次产品的测试数据的鲁棒性满足设定条件时则将当前测试数据插入至先前的训练种群中以形成新的训练种群，以更新得到新的动态测试限值，即MEAT检测方法是根据数据稳健度(适应度/鲁棒性)监控种群进化，实现自适应地对测试限值的动态调整，能够实时有效地检测出存在异常数据的芯片测试数据，从而提高芯片的测试质量。
实施例3
如图11所示，本实施例的产品测试数据的检测系统包括历史数据获取模块1、中间数据获取模块2、分组获取模块3、分布类型获取模块4、第一判断模块5和测试限值获取模块6。
历史数据获取模块1用于获取多个历史批次产品对应的历史测试数据；
在一可实施的方式中，历史测试数据为通过设备规范定义的测试限值的至少六个历史批次产品的测试数据，每个批次产品包括至少30个检测参数；其中，历史测试数据对应的历史批次产品的数量以及每个批次产品中检测参数的数量均可以根据实际情况进行重新确定与调整。
一般将历史测试数据存储为STDF文件，属于批量生产测试数据文件；当然也可以根据实际情况将其存储为其他格式文件。
另外，需要预先在内存中建立静态数据空间，一旦建立静态数据空间就可以开始批量生产测试流程。
在开始检测时，初始化历史测试数据对应的STDF文件，解码出STDF文件对应的STDF内容并存储至静态数据空间；其中，解码得到的STDF内容为ASCII数据。
本实施例的产品测试数据的检测系统还包括数据空间建立模块和存储模块。数据空间建立模块用于预先建立静态数据空间；存储模块用于获取设定格式的历史测试数据，并对历史测试数据进行解码处理并解码后的历史测试数据存储至静态数据空间。
中间数据获取模块2用于对历史测试数据进行筛选处理以获取中间测试数据；
其中，中间数据获取模块2包括筛选单元和剔除单元。
筛选单元用于筛选出历史测试数据中与所有预设测试参数对应的第三测试数据；剔除单元用于从第三测试数据中剔除超出预设测试限值的测试数据以获取中间测试数据。
具体地，基于所有预设测试参数，通过不同的API从静态数据空间输出第三测试数 据。预设测试参数包括但不限于测试项或多个同源性应用程序中站点。
另外，需要通过人工或自动方式检查获取的所有中间测试数据分布情况是否合理，排除明显不合理的测试数据，以保证后期测试限值确定的准确性和可靠性。
分组获取模块3用于根据不同的预设测试参数对中间测试数据进行分组处理以获取多个第一分组；其中，每种预设测试参数对应一个第一分组；
具体地，分组获取模块3包括分组单元和第二判断单元。
分组单元，用于根据不同的预设测试参数对中间测试数据进行分组处理，获取多个中间分组；第二判断单元，用于判断中间分组的大小是否大于或者等于第二设定阈值，若是，则将所述中间分组作为所述第一分组，即剔除包含较少数据量的分组，减少了整体计算量，提高了计算效率，进而提高了整体的检测效率。
分布类型获取模块4用于根据第一分组对应的中间测试数据获取每个第一分组对应的第一分布类型；
第一判断模块5用于判断第一分布类型是否为预设分布类型，若是，则将第一分布类型对应的第一分组作为目标分组；
其中，预设分布类型包括正态分布，即通过对分布类型的比对，将0-1分布等其他分布类型的分组剔除，仅保留正态分布的分组，从而保证了后期测试限值确定的准确性和可靠性。测试限值获取模块6用于根据目标分组对应的中间测试数据获取目标测试限值；
其中，目标测试限值用于对新批次产品的测试数据进行测试。
目标测试限值作为第一个动态限值，保证了开始新批次产品检测之前收紧关键限值，保证了对新批次产品的测试数据的检测准确性。
具体地，测试限值获取模块6包括参数计算单元和测试限值计算单元；
参数计算单元用于根据目标分组对应的中间测试数据计算得到目标分组对应的统计学参数；其中，统计学参数包括平均值和均方差；
测试限值计算单元用于根据统计学参数和预设约束条件(CPK约束条件，即过程能力指数)计算得到测试上限值和测试下限值，并将测试上限值和测试下限值作为目标测试限值。
例如：
Figure PCTCN2021104882-appb-000005
Figure PCTCN2021104882-appb-000006
其中，
Figure PCTCN2021104882-appb-000007
表示平均值，σ表示均方差，CPK为预设约束条件，Dynamic UL表示测试上限值，Dynamic LL表示测试下限值。
在一可实施的方式中，如图2中的(a)所示，仅基于待测产品设计规范得到测试限值为[0，200]；如图2中的(b)所示，当基于统计学参数和CPK约束条件(±CPK*sigma，sigma表示均方差)以计算得到测试上限值和测试下限值[46.13，67.87]，即通过结合CPK约束条件能够计算得到更收紧的测试限值，以保证芯片的测试质量。
本实施例中，基于收集的历史若干批次产品的历史测试数据以及预设测试参数(如测试项)，对其进行过滤、分组等处理，根据每个分组中测试数据的分布类型筛选出目标分组以计算得到测试限值，实现在新批次产品检测之前收紧关键限值，以保证对新批次产品的测试数据的检测准确性，从而提高芯片的测试质量。
实施例4
如图12所示，本实施例的产品测试数据的检测系统是对实施例3的进一步改进，具体地：
检测系统还包括当前数据获取模块7、目标数据获取模块8、第二判断模块9和确定模块10。
当前数据获取模块7用于获取当前批次产品中当前测试组对应的当前测试数据；
目标数据获取模块8用于获取当前测试数据中与不同的预设测试参数对应的多组目标测试数据；
第二判断模块9用于判断目标测试数据是否在对应的目标测试限值内，若是，则确定目标测试数据为正常测试数据；若否，则确定目标测试数据为异常测试数据；
确定模块10用于在设定数量的目标测试数据均为正常测试数据时，则确定当前测试组的当前测试数据通过检测；否则，确定当前测试组的当前测试数据未通过检测。
其中，设定数量的目标测试数据可以为全部数量的目标测试数据，也可以根据实际情况具体确定，例如：当100个目标测试数据中有98个为正常测试数据时则确定当前测试组的当前测试数据通过检测。
在确定当前批次产品中当前测试组对应的当前测试数据通过检测，且预设分布类型为正态分布时，本实施例的检测系统还包括当前种群获取模块11、种群参数计算模块12、第三判断模块13和测试限值更新模块14。
当前种群获取模块11用于将目标分组对应的中间测试数据作为当前训练种群；
种群参数计算模块12用于计算得到当前训练种群对应的初始种群参数；
第三判断模块13用于根据初始种群参数判断当前测试数据是否落入当前训练种群对应的正态分布的中心区域内，若是，则确定当前测试数据的鲁棒性满足预设要求，并将当前测试数据插入至训练种群以形成目标训练种群；
测试限值更新模块14用于根据目标训练种群对应的测试数据更新目标测试限值；
如图5所示，在当前测试数据落入当前训练种群对应的正态分布的中心区域时，则 说明当前测试数据的鲁棒性(适应性)足够强；如图6所示，此时将其插入之前的训练种群中以形成新的训练种群，对应新的统计学参数，从而实现动态建立新的目标测试限值。
其中，在产片批量测试阶段，适应性功能用于连续监视每个芯片的测试数据，随着训练种群不断发展以达到自适应测试的目的。
对于当前批次产品中下一测试组对应的测试数据，所述种群参数计算模块12还用于计算得到所述目标训练种群对应的目标种群参数；
所述第三判断模块13还用于根据所述目标种群参数判断当前测试数据是否落入所述目标训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所述当前测试数据插入至所述目标训练种群以形成新的所述目标训练种群；
其中，对于同一预设测试参数，在当前测试数据中对应的测试数据满足鲁棒性要求时，则将当前测试数据中对应的测试数据插入之前的训练种群中以形成目标训练种群。
具体地，鲁棒性可以通过如下公式确定：
Figure PCTCN2021104882-appb-000008
所述测试限值更新模块14还用于根据新的所述目标训练种群对应的测试数据更新所述目标测试限值。
具体地，根据新的目标训练种群对应的测试数据计算得到对应的统计学参数，并结合CPK约束条件最终计算得到新的目标测试限值。
通过同一批次产品中的新的测试组对应的测试数据及时更新目标测试限值以保证对芯片测试的质量。
即本实施例在量产自动测试过程中，源源不断地将当前芯片的测试数据将作为新个体，通过适应度函数与种群数组进行对比，以评估其鲁棒性。
本实施例的检测方法属于基于进化理论的实时测试数据监控演算法，称为MEAT，该算法结合了静态PAT和动态PAT的特性，并引入了CPK约束条件和演化策略，以实现对消费性芯片的高质量测试。
如图7所示，横轴表示测试数据序列，纵轴Test Data Distribution表示测试数据范围，LL表示测试下限值，UL表示测试上限值，A区域的圆点表示每个当前测试数据；可以得知，基于上述得到的目标测试限值对当前测试数据进行检测。
另外，当预设测试参数包括测试项时，MEAT将每个测试项目作为一个单独的训练种群进行监视；当预设测试参数包括多个同源性应用程序中站点时，则MEAT将多个同源性应用程序中每个站点作为一个单独的训练种群进行监视。其中，如图8所示，对于 多站点应用下的动态限值，每个站点都有独立的限值线。
另外，本实施例的检测系统还包括第四判断模块15；
第四判断模块15，用于判断目标训练种群中的测试数据是否满足预设条件，若满足，则生成第一测试数据以更新目标训练种群；
其中，更新前的目标训练种群与更新后的目标训练种群的统计学参数相差小于第一设定阈值，且更新后的目标训练种群对应的测试数据不满足预设条件；统计学参数包括平均值和均方差。
具体地，第四判断模块15包括四分位数获取单元、第一判断单元和生成单元。
四分位数获取单元用于获取目标训练种群中的测试数据对应的四分位数；
第一判断单元用于判断四分位数中的第一四分位数是否等于第三四分位数，若是，则调用生成单元；
生成单元用于随机生成第一测试数据以更新目标训练种群。
其中，生成单元采用反函数采样方法、Box-Muller变换方法、中央极限定理中的至少一种方式，分别随机生成一组第二测试数据，计算每组第二测试数据对应的统计学参数与更新前的目标训练种群的统计学参数的差值，并选取差值的绝对值最小时对应的第二测试数据作为第一测试数据以更新目标训练种群。当然也可以采用能够随机生成数据的方法来生成测试数据。
采用随机生成的测试数据替换原有的目标训练种群中的测试数据，可以有效地避免种群在进化过程中产生局部收敛以使UL和LL过于接近的情况发生，进而保证动态测试限值的可靠性。
本实施例的检测方法MEAT不需要基于上述内容之外的其他信息，例如晶粒在晶片上的坐标，从而提高了现有的产品检测方式的检测效率及准确度。
下面结合实例具体说明：
如图9所示，为基于现有动态DPAT的方式对测试数据进行检测的检测结果。其中，横轴表示测试数据序列，纵轴表示测试数据范围。DPAT-LL表示测试下限值，DPAT-UL表示测试上限值。在图中的b1处，由于测试数据中出现连续异常数据，DPAT-UL也随之出现明显的抬升，由此可以得知，该检测方式对测试数据具有很高的依赖性，因此连续发布数据则将其检测机制产生较大的影响，甚至可能会失去效力。
如图10所示，为基于MEAT检测方法对测试数据进行检测的检测结果。其中，横轴表示测试数据序列，纵轴表示测试数据范围。MEAT-LL表示测试下限值，MEAT-UL表示测试上限值。可以得知，MEAT检测方法是根据数据稳健度决定是否进行种群进化，降低了MEAT动态限值对测试数据的敏感度，具有更合理的机制来减少对测试数据的依赖性，在该检测方式下甚至可以连续发布数据，在生产过程中能够有效地严格收紧动态 限值。
动态DPAT的检测方式和本实施例的MEAT检测方法的检测方式的检测结果对比数据如下表：
待测芯片数量 DPPM 失效芯片数量 DPAT离群数 MEAT离群数
11440 50874 582 4 515
由上表可以得知，待测芯片数量为11440，失效芯片数量为582，则该批次芯片的DPPM(百万分比的缺陷率)＝(582/11440)*1000000＝50874。
采用现有DPAT的检测方式只能从发布的582个数据中检测出4个异常测试数据，对应的检测率＝(4/582)*100％＝0.69％，而采用本实施例的MEAT离群数能从发布的582个数据中检测出515个异常测试数据，对应的检测率＝(515/582)*100％＝88.49％，由此可知，采用本实施例的测试数据检测方法能够有效地分析出异常测试数据，从而有效地提高了芯片的测试质量。
另外，实验证明，本实施例中的MEAT检测方法对单元测试覆盖率超过C++和Java版本的95％。甚至还可以将MEAT检测方法应用在跨操作系统和编程语言的数据一致性检查方便，以实现可追溯适应性数据的创建和实时存储等。
本实施例中，基于收集的历史若干批次产品的历史测试数据以及预设测试参数(如测试项)，对其进行过滤、分组等处理，根据每个分组中测试数据的分布类型筛选出目标分组以计算得到初始测试限值；当新批次产品的测试数据的鲁棒性满足设定条件时则将当前测试数据插入至先前的训练种群中以形成新的训练种群，以更新得到新的动态测试限值，即MEAT检测方法是根据数据稳健度(适应度/鲁棒性)监控种群进化，实现自适应地对测试限值的动态调整，能够实时有效地检测出存在异常数据的芯片测试数据，从而提高芯片的测试质量。
实施例5
图13为本发明实施例5提供的一种电子设备的结构示意图。电子设备包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，处理器执行程序时实现实施例1或2中任意一实施例中的产品测试数据的检测方法。图13显示的电子设备30仅仅是一个示例，不应对本发明实施例的功能和使用范围带来任何限制。
如图13所示，电子设备30可以以通用计算设备的形式表现，例如其可以为服务器设备。电子设备30的组件可以包括但不限于：上述至少一个处理器31、上述至少一个存储器32、连接不同系统组件(包括存储器32和处理器31)的总线33。
总线33包括数据总线、地址总线和控制总线。
存储器32可以包括易失性存储器，例如随机存取存储器(RAM)321和/或高速缓存存 储器322，还可以进一步包括只读存储器(ROM)323。
存储器32还可以包括具有一组(至少一个)程序模块324的程序/实用工具325，这样的程序模块324包括但不限于：操作系统、一个或者多个应用程序、其它程序模块以及程序数据，这些示例中的每一个或某种组合中可能包括网络环境的实现。
处理器31通过运行存储在存储器32中的计算机程序，从而执行各种功能应用以及数据处理，例如本发明实施例1或2中任意一实施例中的产品测试数据的检测方法。
电子设备30也可以与一个或多个外部设备34(例如键盘、指向设备等)通信。这种通信可以通过输入/输出(I/O)接口35进行。并且，模型生成的设备30还可以通过网络适配器36与一个或者多个网络(例如局域网(LAN)，广域网(WAN)和/或公共网络，例如因特网)通信。如图13所示，网络适配器36通过总线33与模型生成的设备30的其它模块通信。应当明白，尽管图中未示出，可以结合模型生成的设备30使用其它硬件和/或软件模块，包括但不限于：微代码、设备驱动器、冗余处理器、外部磁盘驱动阵列、RAID(磁盘阵列)系统、磁带驱动器以及数据备份存储系统等。
应当注意，尽管在上文详细描述中提及了电子设备的若干单元/模块或子单元/模块，但是这种划分仅仅是示例性的并非强制性的。实际上，根据本发明的实施方式，上文描述的两个或更多单元/模块的特征和功能可以在一个单元/模块中具体化。反之，上文描述的一个单元/模块的特征和功能可以进一步划分为由多个单元/模块来具体化。
实施例6
本实施例提供了一种计算机可读存储介质，其上存储有计算机程序，程序被处理器执行时实现实施例1或2中任意一实施例中的产品测试数据的检测方法中的步骤。
其中，可读存储介质可以采用的更具体可以包括但不限于：便携式盘、硬盘、随机存取存储器、只读存储器、可擦拭可编程只读存储器、光存储器件、磁存储器件或上述的任意合适的组合。
在可能的实施方式中，本发明还可以实现为一种程序产品的形式，其包括程序代码，当程序产品在终端设备上运行时，程序代码用于使终端设备执行实现实施例1或2中任意一实施例中的产品测试数据的检测方法中的步骤。
其中，可以以一种或多种程序设计语言的任意组合来编写用于执行本发明的程序代码，程序代码可以完全地在用户设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户设备上部分在远程设备上执行或完全在远程设备上执行。
虽然以上描述了本发明的具体实施方式，但是本领域的技术人员应当理解，这些仅是举例说明，本发明的保护范围是由所附权利要求书限定的。本领域的技术人员在不背离本发明的原理和实质的前提下，可以对这些实施方式作出多种变更或修改，但这些变更和修改均落入本发明的保护范围。

Claims (13)

  1. 一种产品测试数据的检测方法，其特征在于，所述检测方法包括：
    获取多个历史批次产品对应的历史测试数据；
    对所述历史测试数据进行筛选处理以获取中间测试数据；
    根据不同的预设测试参数对所述中间测试数据进行分组处理以获取多个第一分组；其中，每种所述预设测试参数对应一个所述第一分组；
    根据所述第一分组对应的所述中间测试数据获取每个所述第一分组对应的第一分布类型；
    判断所述第一分布类型是否为预设分布类型，若是，则将所述第一分布类型对应的所述第一分组作为目标分组；
    根据所述目标分组对应的所述中间测试数据获取目标测试限值；
    其中，所述目标测试限值用于对新批次产品的测试数据进行测试。
  2. 如权利要求1所述的产品测试数据的检测方法，其特征在于，所述根据所述目标分组对应的所述中间测试数据获取目标测试限值的步骤包括：
    根据所述目标分组对应的所述中间测试数据计算得到所述目标分组对应的统计学参数；其中，所述统计学参数包括平均值和均方差；
    根据所述统计学参数和预设约束条件计算得到测试上限值和测试下限值，并将所述测试上限值和所述测试下限值作为所述目标测试限值。
  3. 如权利要求1所述的产品测试数据的检测方法，其特征在于，所述根据所述目标分组对应的所述中间测试数据获取目标测试限值的步骤之后还包括：
    获取当前批次产品中当前测试组对应的当前测试数据；
    获取所述当前测试数据中与不同的所述预设测试参数对应的多组目标测试数据；
    判断所述目标测试数据是否在对应的所述目标测试限值内，若是，则确定所述目标测试数据为正常测试数据；若否，则确定所述目标测试数据为异常测试数据；
    在设定数量的所述目标测试数据均为正常测试数据时，则当前测试组的所述当前测试数据通过检测；否则，确定当前测试组的所述当前测试数据未通过检测。
  4. 如权利要求3所述的产品测试数据的检测方法，其特征在于，在确定当前批次产品中当前测试组对应的所述当前测试数据通过检测，且所述预设分布类型为正态分布时，所述检测方法还包括：
    将所述目标分组对应的所述中间测试数据作为当前训练种群；
    计算得到所述当前训练种群对应的初始种群参数；
    根据所述初始种群参数判断所述当前测试数据是否落入所述当前训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所 述当前测试数据插入至所述当前训练种群以形成目标训练种群；
    根据所述目标训练种群对应的测试数据更新所述目标测试限值；
    对于当前批次产品中下一测试组对应的测试数据，计算得到所述目标训练种群对应的目标种群参数；
    根据所述目标种群参数判断当前测试数据是否落入所述目标训练种群对应的正态分布的中心区域内，若是，则确定所述当前测试数据的鲁棒性满足预设要求，并将所述当前测试数据插入至所述目标训练种群以形成新的所述目标训练种群；
    根据新的所述目标训练种群对应的测试数据更新所述目标测试限值。
  5. 如权利要求4所述的产品测试数据的检测方法，其特征在于，所述检测方法还包括：
    判断所述目标训练种群中的测试数据是否满足预设条件，若满足，则生成第一测试数据以更新所述目标训练种群；
    其中，更新前的所述目标训练种群与更新后的所述目标训练种群的统计学参数相差小于第一设定阈值，且更新后的所述目标训练种群对应的测试数据不满足所述预设条件；所述统计学参数包括平均值和均方差。
  6. 如权利要求5所述的产品测试数据的检测方法，其特征在于，所述判断所述目标训练种群中的测试数据是否满足预设条件，若满足，则生成第一测试数据以更新所述目标训练种群的步骤包括：
    获取所述目标训练种群中的测试数据对应的四分位数；
    判断所述四分位数中的第一四分位数是否等于第三四分位数，若是，则随机生成所述第一测试数据以更新所述目标训练种群。
  7. 如权利要求6所述的产品测试数据的检测方法，其特征在于，所述随机生成所述第一测试数据以更新所述目标训练种群的步骤包括：
    采用反函数采样方法、Box-Muller变换方法、中央极限定理中的至少一种方式，分别随机生成一组第二测试数据，计算每组所述第二测试数据对应的统计学参数与更新前的所述目标训练种群的统计学参数的差值，并选取所述差值的绝对值最小时对应的所述第二测试数据作为所述第一测试数据以更新所述目标训练种群。
  8. 如权利要求1所述的产品测试数据的检测方法，其特征在于，所述对所述历史测试数据进行筛选处理以获取中间测试数据的步骤包括：
    筛选出所述历史测试数据中与所有所述预设测试参数对应的第三测试数据；
    从所述第三测试数据中剔除超出预设测试限值的测试数据以获取所述中间测试数据。
  9. 如权利要求1所述的产品测试数据的检测方法，其特征在于，所述根据预设测试参数对所述中间测试数据进行分组处理以获取多个第一分组的步骤包括：
    根据不同的所述预设测试参数对所述中间测试数据进行分组处理，获取多个中间分组；
    判断所述中间分组的大小是否大于或者等于第二设定阈值，若是，则将所述中间分组作为所述第一分组。
  10. 如权利要求8所述的产品测试数据的检测方法，其特征在于，所述获取多个历史批次产品对应的历史测试数据的步骤之前还包括：
    预先建立静态数据空间；
    所述获取多个历史批次产品对应的历史测试数据的步骤之后、所述对所述历史测试数据进行筛选处理以获取中间测试数据的步骤之前包括：
    获取设定格式的所述历史测试数据，并对所述历史测试数据进行解码处理并将解码后的所述历史测试数据存储至所述静态数据空间；
    所述筛选出所述历史测试数据中与所有所述预设测试参数对应的第三测试数据的步骤包括：
    基于所有所述预设测试参数，通过不同的应用程序接口从所述静态数据空间输出所述第三测试数据。
  11. 一种产品测试数据的检测系统，其特征在于，所述检测系统包括：
    历史数据获取模块，用于获取多个历史批次产品对应的历史测试数据；
    中间数据获取模块，用于对所述历史测试数据进行筛选处理以获取中间测试数据；
    分组获取模块，用于根据不同的预设测试参数对所述中间测试数据进行分组处理以获取多个第一分组；其中，每种所述预设测试参数对应一个所述第一分组；
    分布类型获取模块，用于根据所述第一分组对应的所述中间测试数据获取每个所述第一分组对应的第一分布类型；
    第一判断模块，用于判断所述第一分布类型是否为预设分布类型，若是，则将所述第一分布类型对应的所述第一分组作为目标分组；
    测试限值获取模块，用于根据所述目标分组对应的所述中间测试数据获取目标测试限值；
    其中，所述目标测试限值用于对新批次产品的测试数据进行测试。
  12. 一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其特征在于，所述处理器执行计算机程序时实现权利要求1-10中任一项所述的产品测试数据的检测方法。
  13. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1-10中任一项所述的产品测试数据的检测方法的步骤。
PCT/CN2021/104882 2020-09-23 2021-07-07 产品测试数据的检测方法、系统、电子设备和存储介质 WO2022062567A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/119,844 US20230214568A1 (en) 2020-09-23 2023-03-10 Detection method, system, electronic equipment, and storage medium of product test data

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011007511.8 2020-09-23
CN202011007511.8A CN114254261A (zh) 2020-09-23 2020-09-23 产品测试数据的检测方法、系统、电子设备和存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/119,844 Continuation US20230214568A1 (en) 2020-09-23 2023-03-10 Detection method, system, electronic equipment, and storage medium of product test data

Publications (1)

Publication Number Publication Date
WO2022062567A1 true WO2022062567A1 (zh) 2022-03-31

Family

ID=80788799

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/104882 WO2022062567A1 (zh) 2020-09-23 2021-07-07 产品测试数据的检测方法、系统、电子设备和存储介质

Country Status (4)

Country Link
US (1) US20230214568A1 (zh)
CN (1) CN114254261A (zh)
TW (1) TW202212843A (zh)
WO (1) WO2022062567A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114689975A (zh) * 2022-04-01 2022-07-01 深圳市华检测试技术服务有限公司 基于谐波电流的产品测试方法、装置、设备及存储介质
CN114996519A (zh) * 2022-05-17 2022-09-02 苏州佳祺仕信息科技有限公司 数据处理方法、装置、电子设备、存储介质及产品
CN115774185A (zh) * 2023-02-13 2023-03-10 江苏泰治科技股份有限公司 一种车规级芯片dpat检测方法及装置
CN116245256A (zh) * 2023-04-23 2023-06-09 湖州新江浩电子有限公司 一种结合多因素的电容器质量预测方法、系统及存储介质
CN116307289A (zh) * 2023-05-19 2023-06-23 苏州益诚隆纺织科技有限公司 一种纺织品加工工序参数检测预测方法、系统及存储介质

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115020265B (zh) * 2022-07-15 2022-10-11 深圳微迅信息科技有限公司 一种晶圆芯片的检测方法、装置、电子设备及存储介质
CN116381457A (zh) * 2023-04-10 2023-07-04 中山市博测达电子科技有限公司 一种基于数据预统计的光感芯片校正方法
CN117314475B (zh) * 2023-11-28 2024-02-13 德州市水利局水利施工处 一种封堵门防伪溯源过程中异常数据监测方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102117730A (zh) * 2009-12-31 2011-07-06 中芯国际集成电路制造(上海)有限公司 半导体制造过程中的机台参数数据的处理方法和装置
US20130091396A1 (en) * 2011-10-07 2013-04-11 Michele Portolan Packet-based propagation of testing information
CN110138637A (zh) * 2019-05-16 2019-08-16 广东信通通信有限公司 测试数据处理方法、装置、计算机设备和存储介质
CN111177222A (zh) * 2019-12-26 2020-05-19 腾讯云计算（北京）有限责任公司 模型测试方法、装置及计算设备、存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102117730A (zh) * 2009-12-31 2011-07-06 中芯国际集成电路制造(上海)有限公司 半导体制造过程中的机台参数数据的处理方法和装置
US20130091396A1 (en) * 2011-10-07 2013-04-11 Michele Portolan Packet-based propagation of testing information
CN110138637A (zh) * 2019-05-16 2019-08-16 广东信通通信有限公司 测试数据处理方法、装置、计算机设备和存储介质
CN111177222A (zh) * 2019-12-26 2020-05-19 腾讯云计算（北京）有限责任公司 模型测试方法、装置及计算设备、存储介质

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114689975A (zh) * 2022-04-01 2022-07-01 深圳市华检测试技术服务有限公司 基于谐波电流的产品测试方法、装置、设备及存储介质
CN114996519A (zh) * 2022-05-17 2022-09-02 苏州佳祺仕信息科技有限公司 数据处理方法、装置、电子设备、存储介质及产品
CN114996519B (zh) * 2022-05-17 2024-02-09 苏州佳祺仕科技股份有限公司 数据处理方法、装置、电子设备、存储介质及产品
CN115774185A (zh) * 2023-02-13 2023-03-10 江苏泰治科技股份有限公司 一种车规级芯片dpat检测方法及装置
CN116245256A (zh) * 2023-04-23 2023-06-09 湖州新江浩电子有限公司 一种结合多因素的电容器质量预测方法、系统及存储介质
CN116245256B (zh) * 2023-04-23 2023-07-28 湖州新江浩电子有限公司 一种结合多因素的电容器质量预测方法、系统及存储介质
CN116307289A (zh) * 2023-05-19 2023-06-23 苏州益诚隆纺织科技有限公司 一种纺织品加工工序参数检测预测方法、系统及存储介质
CN116307289B (zh) * 2023-05-19 2023-08-18 苏州益诚隆纺织科技有限公司 一种纺织品加工工序参数检测预测方法、系统及存储介质

Also Published As

Publication number Publication date
US20230214568A1 (en) 2023-07-06
TW202212843A (zh) 2022-04-01
CN114254261A (zh) 2022-03-29

Similar Documents

Publication Publication Date Title
WO2022062567A1 (zh) 产品测试数据的检测方法、系统、电子设备和存储介质
WO2022068645A1 (zh) 数据库故障发现方法、装置、电子设备及存储介质
CN110704231A (zh) 一种故障处理方法及装置
CN107992410B (zh) 软件质量监测方法、装置、计算机设备和存储介质
US11632304B2 (en) Methods and systems for characterizing computing system performance using peer-derived performance severity and symptom severity models
WO2014168883A1 (en) System and method for the automatic determination of critical parametric electrical test parameters for inline yield monitoring
US20180173608A1 (en) Software Defect Analysis Tool
WO2023115856A1 (zh) 一种任务异常告警方法及装置
US20180173607A1 (en) Software Defect Detection Tool
CN111626498B (zh) 设备运行状态预测方法、装置、设备及存储介质
CN113837596A (zh) 一种故障确定方法、装置、电子设备及存储介质
CN111143101B (zh) 用于确定故障根源的方法、装置、存储介质及电子设备
CN104364664A (zh) 用于创建、定义和执行spc规则决策树的算法和结构
WO2019019429A1 (zh) 一种虚拟机异常检测方法、装置、设备及存储介质
WO2022143923A1 (zh) 车用电池的健康状况评估方法、系统、电子设备和介质
CN116541222A (zh) 一种硬盘状态数据生成方法、系统、设备及介质
CN114662589A (zh) 一种电表故障研判方法、装置、设备和可读存储介质
CN114328141A (zh) 硬盘故障预警方法及相关组件
KR101542558B1 (ko) 웨이퍼 수율 맵을 분석하는 방법 및 상기 방법을 기록한 기록매체
JP7273669B2 (ja) ストレージシステム及びその制御方法
CN114095394A (zh) 网络节点故障检测方法、装置、电子设备及存储介质
Williams et al. Challenging the paradigm of monitor reduction to achieve lower product costs
CN111666173B (zh) 报错信息处理方法、装置、监控系统和介质
CN116012673A (zh) 晶圆失效模型建立方法、预测方法、装置及相关设备
WO2023057528A1 (en) Ceph failure and verification

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21870946

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21870946

Country of ref document: EP

Kind code of ref document: A1