Electronic component production data management method

Abstract

The invention discloses a data management method based on electronic component production. The method comprises the following steps: acquiring historical defective product data of production of a plurality of electronic components; dividing the historical inferior-quality product data into a plurality of subsequences, and obtaining target inferior-quality product data in the historical inferior-quality product data and the change error degree of the target inferior-quality product data; obtaining initial secondary exponential smoothing coefficients of different target defective data, and adjusting the initial secondary exponential smoothing coefficients of all the target defective data by using change significance degree differences and change error degree differences between different target defective data to obtain new secondary exponential smoothing coefficients; according to the new secondary exponential smoothing coefficient, electronic component defective product prediction is carried out, and defective product data is written into a block chain; according to the method, in the electronic component generation process, electronic component defective product prediction and production defective product data writing into the block chain can be carried out by generating the new secondary exponential smoothing coefficient, and the management efficiency and traceability efficiency of production data are improved.

Description

Technical Field

[0001] The present invention belongs to the field of electronic components, and specifically refers to a method for managing production data based on electronic components. Background Art

[0002] Electronic components are the foundation of the modern electronics industry. They are the basic components of electronic equipment. With the rapid development of electronic technology, the production process of electronic components has become more and more complicated, involving many processes and parameters. Effective production data management is of great significance to ensure product quality, improve production efficiency and reduce costs.

[0003] There is a need to store and manage the defective data generated in each production link. At present, when storing defective data, the conventional Huffman coding algorithm is usually used for direct compression storage. However, there is an accumulation of defective errors in the manufacturing process of electronic components, which leads to a certain inaccuracy in the defective probability distribution of each link in the defective data, making the accuracy of the defective probability distribution data low. The efficiency of tracing back after compressed storage is very low, making the management of production data arduous and inefficient. It is urgent to find a better data management method for traceability to replace the traditional compression storage management method for traceability. Summary of the invention

[0004] In order to solve the problems in the above-mentioned prior art that the accumulation of defective errors in the manufacturing process of electronic components leads to a certain inaccuracy in the defective probability distribution of each link in the defective data, making the accuracy of the defective probability distribution data low, and the traceability efficiency after compression and storage is very low, making the management of production data cumbersome and inefficient, the present invention proposes a production data management method based on electronic components to improve the above-mentioned problems.

[0005] The specific application is as follows:

[0006] A method for managing production data of electronic components, comprising the following steps:

[0007] Obtain historical defective data of the production of several electronic components, where any data point in the historical defective data corresponds to a production time sample point sequence and a mean defective probability under the production time sample point sequence;

[0008] Divide the historical defective data into several subsequences, and obtain the target defective data and the degree of variation error of the target defective data in the several historical defective data according to the degree of variation between adjacent subsequences in the historical defective data and the length of the subsequences, wherein the length is the number of data points multiplied by the number length of the production time sample point sequence, and the degree of variation indicates the fluctuation of the defective probability mean between adjacent subsequences;

[0009] Obtaining initial quadratic exponential smoothing coefficients of different target defective data, and adjusting the initial quadratic exponential smoothing coefficients of all target defective data by using the difference in the degree of change significance and the degree of change error between the different target defective data, to obtain new quadratic exponential smoothing coefficients, wherein the quadratic exponential smoothing coefficients include a horizontal smoothing coefficient and a trend smoothing coefficient;

[0010] The defective production of electronic components is predicted based on the new quadratic exponential smoothing coefficient and the defective production data is written into the blockchain.

[0011] Furthermore, the historical defective product data is divided into several subsequences, including the following specific methods:

[0012] A two-dimensional coordinate system is constructed, and the production time sample point sequence of the data points in the historical defective data is taken as the horizontal axis of the two-dimensional coordinate system, and the defective probability mean corresponding to the production time sample point sequence of the data points in the historical defective data is taken as the vertical axis of the two-dimensional coordinate system. The K-Means++ clustering algorithm is used to cluster the data points of all historical defective data in the two-dimensional coordinate system to obtain several clustering clusters, and the sequences formed by all the data points of each historical defective data in the clustering cluster are respectively recorded as subsequences of each historical defective data.

[0013] Furthermore, the method of obtaining target defective data and the degree of error of the target defective data in a number of historical defective data according to the degree of change between adjacent subsequences in the historical defective data and the length of the subsequence includes the following specific methods:

[0014] Obtain target defective product data according to the significant degree of changes between adjacent subsequences in historical defective product data;

[0015] The ratio between the change significance of the subsequence and the change significance of the mean subsequence under the subsequence corresponding cluster is recorded as the change significance coefficient of the subsequence, and the sequence formed by the change significance coefficients of all subsequences in the target defective data is recorded as the change significance sequence of the target defective data;

[0016] The specific calculation method of the variation error degree of the target defective data is as follows:

[0017]

[0018] Among them, GH represents the degree of error of the target defective data; n represents the number of significant degree coefficients of the target defective data in the significant degree sequence of the change that meet the preset threshold; G r It represents the rth significant change coefficient in the significant change sequence corresponding to the target defective data; Tanh represents the normalization function.

[0019] Furthermore, the mean subsequence includes the following specific methods:

[0020] The TWED algorithm is used to dynamically time-length-regularize all subsequences in any clustering cluster to make the lengths of all subsequences in the clustering cluster the same. All subsequences with the same length in the clustering cluster are recorded as target subsequences, and the average defective probability value of the data points at the same position in all target subsequences is obtained. The sequence formed by the average defective probability values ​​of the data points at the same position in the clustering cluster is recorded as the mean subsequence under the corresponding clustering cluster.

[0021] Furthermore, the target defective product data is obtained according to the degree of change between adjacent subsequences in the historical defective product data, including the specific method of:

[0022] The TWED algorithm is used to obtain the TWED change significance between the adjacent subsequences in the cluster and the mean subsequence under the subsequence corresponding cluster, and the change significance parameter of the subsequence is recorded as the cumulative value of the change significance parameter of all subsequences of the adjacent historical defective data as the change significance factor D of the historical defective data, and exp(-D) is recorded as the confidence of the historical defective data, where exp() represents an exponential function with a natural constant as the base;

[0023] The historical defective product data with a confidence level greater than a preset confidence level threshold is recorded as the target defective product data.

[0024] Furthermore, the specific method of obtaining the initial quadratic exponential smoothing coefficient of different target defective product data includes:

[0025] The quadratic exponential smoothing DES algorithm is used to obtain the quadratic exponential smoothing coefficient of any target defective data, which is recorded as the initial quadratic exponential smoothing coefficient of the target defective data. The quadratic exponential smoothing DES algorithm formula is:

[0026] Horizontal smoothing formula: S t =αY t +(1-α)S t-1 ;

[0027] Among them, S t represents the horizontal smoothing value at time t, Y t represents the actual observed value at time t, S t-1 represents the horizontal smoothing value at time t-1, and α represents the horizontal smoothing coefficient;

[0028] Trend smoothing formula: T t =β(S t -S t-1 )-)+(1-β)T t-1 ;

[0029] Among them, Tt Represents the trend smoothing value at time t, T t-1 represents the trend smoothing value at time t-1, and β represents the trend smoothing coefficient;

[0030] The quadratic exponential smoothing coefficient includes a horizontal smoothing coefficient α and a trend smoothing coefficient β.

[0031] Furthermore, the initial quadratic exponential smoothing coefficients of all target defective data are adjusted by using the difference in the degree of change significance and the difference in the degree of change error between different target defective data to obtain new quadratic exponential smoothing coefficients, including the specific method of:

[0032] The matching degree of the target defective product data is obtained according to the difference in the degree of change significance and the degree of change error between different target defective product data;

[0033] The specific calculation method of the new quadratic exponential smoothing coefficient is:

[0034]

[0035] in, represents the quadratic exponential smoothing coefficient; G i represents the initial quadratic exponential smoothing coefficient of the i-th target defect data, PG i represents the matching degree of the i-th target defective product data; m represents the number of target defective product data; the quadratic exponential smoothing coefficient realizes the conversion between the horizontal smoothing coefficient α and the trend smoothing coefficient β through matrix encoding and decoding.

[0036] Furthermore, the specific method of obtaining the matching degree of the target defective product data according to the difference in the degree of change significance and the degree of change error between different target defective product data includes:

[0037] Obtain the difference factors between the target defective product data according to the difference in the degree of change significance and the degree of change error between different target defective product data;

[0038] Will is recorded as the matching degree of the i-th target defective data; m represents the number of target defective data; Cy ij It represents the difference factor between the i-th target defective data and the j-th target defective data other than the i-th target defective data; L1norm() represents the L1 normalization function.

[0039] Furthermore, the specific method of obtaining the difference factor between the target defective data according to the difference in the degree of change significance and the difference in the degree of change error between different target defective data includes:

[0040] The absolute value of the difference between the change significance of the i-th target defective data and the change significance of the j-th target defective data excluding the i-th target defective data is recorded as the first comparison difference; the absolute value of the difference between the change error degree between the i-th target defective data and the j-th target defective data excluding the i-th target defective data is recorded as the second comparison difference;

[0041] The product of the first comparison difference and the second comparison difference is recorded as the difference factor between the i-th target defective data and the j-th target defective data other than the i-th target defective data.

[0042] Furthermore, the specific method of predicting the production of defective electronic components and writing the production defective data into the blockchain according to the new quadratic exponential smoothing coefficient is as follows:

[0043] The sequence of defective probability values ​​of electronic components in all production links is recorded as the first batch of defective data. The first batch of product data is traversed by sliding a sliding window of a preset length. The defective prediction value of the next data point of the last data point in the sliding window is obtained by combining the new quadratic exponential smoothing coefficient and using the quadratic exponential smoothing DES algorithm. Exp[-|KK′|] is recorded as the defective prediction probability index ratio of the data points in the first batch of product data, where K represents the defective probability mean of the data points in the first batch of product data, and K′ represents the predicted defective probability mean of the data points in the first batch of product data. The defective probability mean corresponding to all data points in the first batch of product data whose probability index ratio is greater than a preset predicted probability index ratio threshold is set to 0, and the data points whose probability index ratio is less than or equal to the preset predicted probability index ratio threshold remains unchanged, so as to obtain the second batch of product data, and the second batch of product data is written into the public blockchain for chain operation. The second batch of product data at least includes the time of occurrence of defective products, the link where the defective products occur, and the environmental factors where the defective products occur; the environmental factors at least include the current working temperature, working humidity, working vibration intensity, working current intensity, and working voltage intensity.

[0044] The beneficial effects of the electronic component production data management method of the present invention are as follows:

[0045] The present invention obtains historical defective data of the production of several electronic components, wherein any data point in the historical defective data corresponds to a production time sample point sequence and a defective probability mean under the production time sample point sequence; divides the historical defective data into several subsequences, and obtains target defective data and the degree of change error of the target defective data in the historical defective data according to the degree of change significance between adjacent subsequences in the historical defective data and the length of the subsequences; obtains initial quadratic exponential smoothing coefficients of different target defective data, and adjusts the initial quadratic exponential smoothing coefficients of all target defective data by using the difference in the degree of change significance and the difference in the degree of change error between different target defective data to obtain new quadratic exponential smoothing coefficients; predicts the production of defective electronic components and writes the production defective data into a blockchain according to the new quadratic exponential smoothing coefficients; the present invention can predict the production of defective electronic components and write the production defective data into a blockchain during the generation process of electronic components by generating new quadratic exponential smoothing coefficients, thereby improving the management efficiency and traceability efficiency of production data. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a flowchart of a method for managing production data of electronic components. DETAILED DESCRIPTION

[0047] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example 1

[0049] Figure 1 The present invention is a flowchart of a method for managing production data of electronic components.

[0050] The method comprises steps S1-S4:

[0051] S1: Obtain historical defective data of the production of several electronic components. Any data point in the historical defective data corresponds to a production time sample point sequence and a mean defective probability under the production time sample point sequence;

[0052] It should be noted that when predicting the defective rate of electronic components, due to the influence of wafer processing technology, environmental factors and production materials, there are accumulated errors in the defective rate in multiple production links during the production process of electronic components. For example, under normal circumstances, the production links of electronic components include: material preparation, wafer processing, oxidation, lithography, etching, thin film deposition, interconnection, cleaning and anti-corrosion treatment, quality inspection and packaging. Improper material selection in the material preparation link will lead to defective errors. In addition, if the material quality cannot be effectively controlled, it may cause defective problems; if the pressure, temperature or time and other parameters are not accurately controlled in the wafer processing, oxidation and lithography links, it will cause problems with the structural tightness and defective rate of the components. The accuracy of the inspection method or equipment in the quality inspection link will affect the quality assessment results of the electronic components, and then affect the product quality and defective rate; if the packaging materials are improperly selected or the packaging method is unreasonable in the packaging link, the electronic components will be damaged or lost during transportation, increasing the defective rate;

[0053] In the multiple production links of the above-mentioned electronic components, there is a certain difference between the actual defective rate of a production link and the defective rate specified in the production link, which is the defective rate error. When the defective rate error increases gradually in multiple production links, the defective rate error is accumulated.

[0054] Therefore, this application analyzes the changing characteristics of historical defective data, optimizes the quadratic exponential smoothing coefficient, and obtains accurate defective rate prediction results, so as to improve the efficiency of blockchain operation and traceability of defective data, accurately filter out erroneous prediction results, and improve management efficiency;

[0055] It should be noted that the present application can more accurately reflect the average probability of defective products within a time period by dividing the data point into a sequence of production time sample points, rather than predicting the probability of defective products at a time point and using the average value of the probability of defective products within the time period to describe the probability of defective products at a data point.

[0056] S2: Divide the historical defective data into several subsequences, and obtain the target defective data and the degree of variation error of the target defective data in the several historical defective data according to the degree of variation between adjacent subsequences in the historical defective data and the length of the subsequences, wherein the length is the number of data points multiplied by the number length of the production time sample point sequence, and the degree of variation indicates the fluctuation of the defective probability mean between adjacent subsequences;

[0057] It should be noted that by constructing a two-dimensional coordinate system and using the K-Means++ clustering algorithm to cluster all historical defective data, all subsequences in the obtained clusters reflect the similar defective rate change trends of the included data points in the corresponding production links, that is, each cluster represents a change trend, which can more clearly understand the changes in the defective rate in different production links, facilitate the subsequent acquisition of the defective rate change trend of the corresponding production link, and better predict the defective rate.

[0058] S3: obtaining initial quadratic exponential smoothing coefficients of different target defective data, and adjusting the initial quadratic exponential smoothing coefficients of all target defective data by using the difference in the degree of change significance and the degree of change error between the different target defective data to obtain new quadratic exponential smoothing coefficients, wherein the quadratic exponential smoothing coefficients include a horizontal smoothing coefficient and a trend smoothing coefficient;

[0059] S4: Predict the production of defective electronic components based on the new quadratic exponential smoothing coefficient and write the production defective data into the blockchain.

[0060] Furthermore, the historical defective product data is divided into several subsequences, including the following specific methods:

[0061] A two-dimensional coordinate system is constructed, and the production time sample point sequence of the data points in the historical defective data is taken as the horizontal axis of the two-dimensional coordinate system, and the defective probability mean corresponding to the production time sample point sequence of the data points in the historical defective data is taken as the vertical axis of the two-dimensional coordinate system. The K-Means++ clustering algorithm is used to cluster the data points of all historical defective data in the two-dimensional coordinate system to obtain several clustering clusters, and the sequences formed by all the data points of each historical defective data in the clustering cluster are respectively recorded as subsequences of each historical defective data.

[0062] Furthermore, the method of obtaining target defective data and the degree of error of the target defective data in a number of historical defective data according to the degree of change between adjacent subsequences in the historical defective data and the length of the subsequence includes the following specific methods:

[0063] Obtain target defective product data according to the significant degree of changes between adjacent subsequences in historical defective product data;

[0064] The ratio between the change significance of the subsequence and the change significance of the mean subsequence under the subsequence corresponding cluster is recorded as the change significance coefficient of the subsequence, and the sequence formed by the change significance coefficients of all subsequences in the target defective data is recorded as the change significance sequence of the target defective data;

[0065] The specific calculation method of the variation error degree of the target defective data is as follows:

[0066]

[0067] Among them, GH represents the degree of error of the target defective data; n represents the number of significant degree coefficients of the target defective data in the significant degree sequence of the change that meet the preset threshold; G r It represents the rth significant change coefficient in the significant change sequence corresponding to the target defective data; Tanh represents the normalization function.

[0068] Furthermore, the mean subsequence includes the following specific methods:

[0069] The TWED algorithm is used to dynamically time-length-regularize all subsequences in any clustering cluster to make the lengths of all subsequences in the clustering cluster the same. All subsequences with the same length in the clustering cluster are recorded as target subsequences, and the average defective probability value of the data points at the same position in all target subsequences is obtained. The sequence formed by the average defective probability values ​​of the data points at the same position in the clustering cluster is recorded as the mean subsequence under the corresponding clustering cluster.

[0070] It should be noted that the TWED (Time Warp Edit Distance) algorithm is an elastic distance measurement method for comparing time series. Since the TWED algorithm is an existing algorithm, it will not be described in detail in this embodiment.

[0071] Furthermore, the target defective product data is obtained according to the degree of change between adjacent subsequences in the historical defective product data, including the specific method of:

[0072] The TWED algorithm is used to obtain the TWED change significance between the adjacent subsequences in the cluster and the mean subsequence under the subsequence corresponding cluster, and the change significance parameter of the subsequence is recorded as the cumulative value of the change significance parameter of all subsequences of the adjacent historical defective data as the change significance factor D of the historical defective data, and exp(-D) is recorded as the confidence of the historical defective data, where exp() represents an exponential function with a natural constant as the base;

[0073] The historical defective product data with a confidence level greater than a preset confidence level threshold is recorded as the target defective product data.

[0074] Furthermore, the specific method of obtaining the initial quadratic exponential smoothing coefficient of different target defective product data includes:

[0075] The quadratic exponential smoothing DES algorithm is used to obtain the quadratic exponential smoothing coefficient of any target defective data, which is recorded as the initial quadratic exponential smoothing coefficient of the target defective data. The quadratic exponential smoothing DES algorithm formula is:

[0076] Horizontal smoothing formula: S t =αYt +(1-α)S t-1 ;

[0077] Among them, S t represents the horizontal smoothing value at time t, Y t represents the actual observed value at time t, S t-1 represents the horizontal smoothing value at time t-1, and α represents the horizontal smoothing coefficient;

[0078] Trend smoothing formula: T t =β(S t -S t-1 )-)+(1-β)T t-1 ;

[0079] Among them, T t Represents the trend smoothing value at time t, T t-1 represents the trend smoothing value at time t-1, and β represents the trend smoothing coefficient;

[0080] The quadratic exponential smoothing coefficient includes a horizontal smoothing coefficient α and a trend smoothing coefficient β.

[0081] Furthermore, the initial quadratic exponential smoothing coefficients of all target defective data are adjusted by using the difference in the degree of change significance and the difference in the degree of change error between different target defective data to obtain new quadratic exponential smoothing coefficients, including the specific method of:

[0082] The matching degree of the target defective product data is obtained according to the difference in the degree of change significance and the degree of change error between different target defective product data;

[0083] The specific calculation method of the new quadratic exponential smoothing coefficient is:

[0084]

[0085] in, represents the quadratic exponential smoothing coefficient; G i represents the initial quadratic exponential smoothing coefficient of the i-th target defect data, PG i represents the matching degree of the i-th target defective product data; m represents the number of target defective product data; the quadratic exponential smoothing coefficient realizes the conversion between the horizontal smoothing coefficient α and the trend smoothing coefficient β through matrix encoding and decoding.

[0086] It should be noted that the matching degree of the target defective data reflects the distribution characteristics of the corresponding change error degree of the target defective data, and is obtained based on the difference in change significance and change error degree between the target defective data and other target defective data, reflecting the similarity between the target defective data and other target defective data. The greater the matching degree of the target defective data, the smaller the difference in change significance and change error degree between the target defective data and all other target defective data, and the more similar the change trends between the target defective data and other target defective data. Therefore, when the target defective data is used for defective rate prediction, the greater the effect on improving the accuracy of the defective rate prediction results.

[0087] Furthermore, the specific method of obtaining the matching degree of the target defective product data according to the difference in the degree of change significance and the degree of change error between different target defective product data includes:

[0088] Obtain the difference factors between the target defective product data according to the difference in the degree of change significance and the degree of change error between different target defective product data;

[0089] Will is recorded as the matching degree of the i-th target defective data; m represents the number of target defective data; Cy ij It represents the difference factor between the i-th target defective data and the j-th target defective data other than the i-th target defective data; L1norm() represents the L1 normalization function.

[0090] Furthermore, the specific method of obtaining the difference factor between the target defective data according to the difference in the degree of change significance and the difference in the degree of change error between different target defective data includes:

[0091] The absolute value of the difference between the change significance of the i-th target defective data and the change significance of the j-th target defective data excluding the i-th target defective data is recorded as the first comparison difference; the absolute value of the difference between the change error degree of the i-th target defective data and the j-th target defective data excluding the i-th target defective data is recorded as the second comparison difference;

[0092] The product of the first comparison difference and the second comparison difference is recorded as the difference factor between the i-th target defective data and the j-th target defective data other than the i-th target defective data.

[0093] Furthermore, the specific method of predicting the production of defective electronic components and writing the production defective data into the blockchain according to the new quadratic exponential smoothing coefficient is as follows:

[0094] The sequence of defective probability values ​​of electronic components in all production links is recorded as the first batch of defective data. The first batch of product data is traversed by sliding a sliding window of a preset length. The defective prediction value of the next data point of the last data point in the sliding window is obtained by combining the new quadratic exponential smoothing coefficient and using the quadratic exponential smoothing DES algorithm. Exp[-|KK′|] is recorded as the defective prediction probability index ratio of the data points in the first batch of product data, where K represents the defective probability mean of the data points in the first batch of product data, and K′ represents the predicted defective probability mean of the data points in the first batch of product data. The defective probability mean corresponding to all data points in the first batch of product data whose probability index ratio is greater than a preset predicted probability index ratio threshold is set to 0, and the data points whose probability index ratio is less than or equal to the preset predicted probability index ratio threshold remains unchanged, so as to obtain the second batch of product data, and the second batch of product data is written into the public blockchain for chain operation. The second batch of product data at least includes the time of occurrence of defective products, the link where the defective products occur, and the environmental factors where the defective products occur; the environmental factors at least include the current working temperature, working humidity, working vibration intensity, working current intensity, and working voltage intensity.

[0095] It should be noted that the length of the sliding window and the threshold of the predicted probability index ratio are preset to 4 and 0.6 respectively based on experience, and can be adjusted according to actual conditions, and are not specifically limited in this embodiment.

[0096] Specifically, this embodiment writes defective product data into the blockchain for subsequent traceability operations, which will not be elaborated on. It only uses blockchain traceability to compare with traditional data compression for traceability, thereby improving the efficiency of defective product data traceability and improving the efficiency of production management data.

[0097] The present invention and its embodiments are described above, which is not restrictive. What is shown in the accompanying drawings is only one of the embodiments of the present invention, and the actual content is not limited to this. In short, if ordinary technicians in this field are inspired by it and do not deviate from the purpose of the invention, they can creatively design structural methods and embodiments similar to the technical scheme, which should all fall within the scope of protection of the present invention.

Claims

1. A method for managing production data of electronic components, characterized in that: The following steps are involved: Obtain historical defective data of the production of several electronic components, where any data point in the historical defective data corresponds to a production time sample point sequence and a mean defective probability under the production time sample point sequence; Divide the historical defective data into several subsequences, and obtain the target defective data and the degree of variation error of the target defective data in the several historical defective data according to the degree of variation between adjacent subsequences in the historical defective data and the length of the subsequences, wherein the length is the number of data points multiplied by the number length of the production time sample point sequence, and the degree of variation indicates the fluctuation of the defective probability mean between adjacent subsequences; Obtaining initial quadratic exponential smoothing coefficients of different target defective data, and adjusting the initial quadratic exponential smoothing coefficients of all target defective data by using the difference in the degree of change significance and the degree of change error between the different target defective data, to obtain new quadratic exponential smoothing coefficients, wherein the quadratic exponential smoothing coefficients include a horizontal smoothing coefficient and a trend smoothing coefficient; The defective production of electronic components is predicted based on the new quadratic exponential smoothing coefficient and the defective production data is written into the blockchain.

2. The method for managing electronic component production data according to claim 1, characterized in that: The specific method of dividing the historical defective product data into several subsequences includes: A two-dimensional coordinate system is constructed, and the production time sample point sequence of the data points in the historical defective data is taken as the horizontal axis of the two-dimensional coordinate system, and the defective probability mean corresponding to the production time sample point sequence of the data points in the historical defective data is taken as the vertical axis of the two-dimensional coordinate system. The K-Means++ clustering algorithm is used to cluster the data points of all historical defective data in the two-dimensional coordinate system to obtain several clustering clusters, and the sequences formed by all the data points of each historical defective data in the clustering cluster are respectively recorded as subsequences of each historical defective data.

3. The method for managing electronic component production data according to claim 2, characterized in that: The specific method of obtaining target defective data and the degree of error of the target defective data in a number of historical defective data according to the degree of change between adjacent subsequences in the historical defective data and the length of the subsequences includes: Obtain target defective product data according to the significant degree of changes between adjacent subsequences in historical defective product data; The ratio between the change significance of the subsequence and the change significance of the mean subsequence under the subsequence corresponding cluster is recorded as the change significance coefficient of the subsequence, and the sequence formed by the change significance coefficients of all subsequences in the target defective data is recorded as the change significance sequence of the target defective data; The specific calculation method of the variation error degree of the target defective data is as follows: Among them, GH represents the degree of error of the target defective data; n represents the number of significant degree coefficients of the target defective data in the significant degree sequence of the change that meet the preset threshold; G r It represents the rth significant change coefficient in the significant change sequence corresponding to the target defective data; Tanh represents the normalization function.

4. The method for managing electronic component production data according to claim 3, characterized in that: The mean subsequence includes the following specific methods: The TWED algorithm is used to dynamically time-length-regularize all subsequences in any clustering cluster to make the lengths of all subsequences in the clustering cluster the same. All subsequences with the same length in the clustering cluster are recorded as target subsequences, and the average defective probability value of the data points at the same position in all target subsequences is obtained. The sequence formed by the average defective probability values ​​of the data points at the same position in the clustering cluster is recorded as the mean subsequence under the corresponding clustering cluster.

5. The method for managing production data of electronic components according to claim 4, characterized in that: The specific method of obtaining the target defective product data according to the significant degree of change between adjacent subsequences in the historical defective product data is as follows: The TWED algorithm is used to obtain the TWED change significance between the adjacent subsequences in the cluster and the mean subsequence under the subsequence corresponding cluster, and the change significance parameter of the subsequence is recorded as the cumulative value of the change significance parameter of all subsequences of the adjacent historical defective data as the change significance factor D of the historical defective data, and exp(-D) is recorded as the confidence of the historical defective data, where exp() represents an exponential function with a natural constant as the base; The historical defective product data with a confidence level greater than a preset confidence level threshold is recorded as the target defective product data.

6. The method for managing electronic component production data according to claim 5, characterized in that: The specific method of obtaining the initial quadratic exponential smoothing coefficients of different target defective product data includes: The quadratic exponential smoothing DES algorithm is used to obtain the quadratic exponential smoothing coefficient of any target defective data, which is recorded as the initial quadratic exponential smoothing coefficient of the target defective data. The quadratic exponential smoothing DES algorithm formula is: Horizontal smoothing formula: S t =αY t +(1-α)S t-1 ; Among them, S t represents the horizontal smoothing value at time t, Y t represents the actual observed value at time t, S t-1 represents the horizontal smoothing value at time t-1, and α represents the horizontal smoothing coefficient; Trend smoothing formula: T t =β(S t -S t-1 )-)+(1-β)T t-1 ; Among them, T t Represents the trend smoothing value at time t, T t-1 represents the trend smoothing value at time t-1, and β represents the trend smoothing coefficient; The quadratic exponential smoothing coefficient includes a horizontal smoothing coefficient α and a trend smoothing coefficient β.

7. The method for managing electronic component production data according to claim 6, characterized in that: The method of adjusting the initial quadratic exponential smoothing coefficients of all target defective data by utilizing the difference in the degree of change significance and the degree of change error between different target defective data to obtain a new quadratic exponential smoothing coefficient includes the following specific methods: The matching degree of the target defective product data is obtained according to the difference in the degree of change significance and the degree of change error between different target defective product data; The specific calculation method of the new quadratic exponential smoothing coefficient is: in, represents the quadratic exponential smoothing coefficient; G i represents the initial quadratic exponential smoothing coefficient of the i-th target defect data, PG i represents the matching degree of the i-th target defective product data; m represents the number of target defective product data; the quadratic exponential smoothing coefficient realizes the conversion between the horizontal smoothing coefficient α and the trend smoothing coefficient β through matrix encoding and decoding.

8. The method for managing electronic component production data according to claim 7, characterized in that: The specific method of obtaining the matching degree of the target defective product data according to the difference in the degree of change significance and the degree of change error between different target defective product data includes: Obtain the difference factors between the target defective product data according to the difference in the degree of change significance and the degree of change error between different target defective product data; Will is recorded as the matching degree of the i-th target defective data; m represents the number of target defective data; Cy ij It represents the difference factor between the i-th target defective data and the j-th target defective data other than the i-th target defective data; L1norm() represents the L1 normalization function.

9. The method for managing electronic component production data according to claim 8, characterized in that: The specific method of obtaining the difference factor between the target defective product data according to the difference in the degree of change significance and the degree of change error between different target defective product data includes: The absolute value of the difference between the change significance of the i-th target defective data and the change significance of the j-th target defective data excluding the i-th target defective data is recorded as the first comparison difference; the absolute value of the difference between the change error degree of the i-th target defective data and the j-th target defective data excluding the i-th target defective data is recorded as the second comparison difference; The product of the first comparison difference and the second comparison difference is recorded as the difference factor between the i-th target defective data and the j-th target defective data other than the i-th target defective data.

10. The method for managing electronic component production data according to claim 9, characterized in that: The specific method of predicting the production of defective electronic components and writing the production defective data into the blockchain according to the new quadratic exponential smoothing coefficient is as follows: The sequence of defective probability values ​​of electronic components in all production links is recorded as the first batch of defective data. The first batch of product data is traversed by sliding a sliding window of a preset length. The defective prediction value of the next data point of the last data point in the sliding window is obtained by combining the new quadratic exponential smoothing coefficient and using the quadratic exponential smoothing DES algorithm. Exp[-|KK′|] is recorded as the defective prediction probability index ratio of the data points in the first batch of product data, where K represents the defective probability mean of the data points in the first batch of product data, and K′ represents the predicted defective probability mean of the data points in the first batch of product data. The defective probability mean corresponding to all data points in the first batch of product data whose probability index ratio is greater than a preset predicted probability index ratio threshold is set to 0, and the data points whose probability index ratio is less than or equal to the preset predicted probability index ratio threshold remains unchanged, so as to obtain the second batch of product data, and the second batch of product data is written into the public blockchain for chain operation. The second batch of product data at least includes the time of occurrence of defective products, the link where the defective products occur, and the environmental factors where the defective products occur; the environmental factors at least include the current working temperature, working humidity, working vibration intensity, working current intensity, and working voltage intensity.

Images