Photoelectric product production management system and method based on Internet of Things

Abstract

The invention discloses a photoelectric product production management system and method based on the Internet of Things, and relates to the technical field of the Internet of Things, and the system comprises a preprocessing module, a feature engineering module and a classification prediction module. The preprocessing module is used for arranging a sensor to collect multiple batches of data and manually marking equipment states corresponding to different batches to obtain the product qualification rate; the feature engineering module is used for performing standardization processing on the acquired data, constructing a regression model by taking the acquired data as an independent variable and taking a product qualification rate as a dependent variable, and screening the independent variable as a feature according to an independent variable coefficient of the regression model; and the classification prediction module is used for constructing a classification model by taking the screened independent variable as a feature and combining a labeling condition, and performing equipment state prediction on newly collected data. The invention also provides a method for realizing the system. The method can effectively improve the condition that the core factors influencing the product quality and the production efficiency are difficult to accurately identify in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to an Internet of Things-based optoelectronic product production management system and method. Background Art

[0002] The production of optoelectronic products is a highly sophisticated and complex process, covering multiple links from raw material processing, parts manufacturing to product assembly, and each link requires high-precision equipment and strict process control. Production management plays a key role in this process, involving the maintenance and monitoring of production equipment, the scheduling and optimization of production processes, the testing and assurance of product quality, and the coordination and management of personnel.

[0003] In traditional optoelectronic product production management technology, most of them rely on manual inspections to monitor equipment status. This method is not only manpower-consuming, but also has a long inspection time interval. It is impossible to obtain equipment operation data in real time and continuously, which makes it difficult to capture instantaneous anomalies and subtle changes in equipment operation. In addition, due to the lack of deep mining and analysis capabilities of production equipment operation data, it is difficult to extract valuable information from massive production data, and it is impossible to accurately identify the core factors affecting product quality and production efficiency. The optimization and improvement of the production process often rely on experience judgment and lack scientific basis and pertinence. Summary of the invention

[0004] The purpose of the present invention is to provide an optoelectronic product production management system and method based on the Internet of Things to solve the problems raised in the prior art.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an optoelectronic product production management system based on the Internet of Things, the system comprising a preprocessing module, a feature engineering module and a classification prediction module; The preprocessing module is used to arrange sensors to collect multiple batches of data and manually mark the equipment status corresponding to different batches to obtain the product qualification rate; The feature engineering module is used to standardize the acquired data, and to construct a regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable, and to select the independent variable as the feature according to the independent variable coefficient of the regression model; The classification prediction module is used to construct a classification model based on the screened independent variables as features and combined with manual annotation to standardize the newly collected data and predict the equipment status; The output end of the preprocessing module is connected to the input end of the feature engineering module; the output end of the feature engineering module is connected to the input end of the classification prediction module.

[0006] The preprocessing module includes a data acquisition unit, a manual labeling unit and a product reporting unit; The data acquisition unit is used to arrange sensors on optoelectronic product production equipment and collect multiple batches of data; The manual labeling unit is used to manually label the device status of the collected multiple batches of data; The product reporting unit is used to obtain the production qualification rate of the optoelectronic products according to the production report of each batch of products; The output end of the data acquisition unit is connected to the input end of the manual labeling unit; the output end of the manual labeling unit is connected to the input end of the product reporting unit; the output end of the product reporting unit is connected to the input end of the feature engineering module.

[0007] The feature engineering module includes a data standardization unit, a regression model unit and a variable screening unit; The data standardization unit is used to perform Z-Score standardization processing on the acquired data; The regression model unit is used to construct a polynomial regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable; The variable screening unit is used to calculate the t-statistic of the coefficient of the independent variable of the polynomial regression model and screen the independent variable as a feature according to the preset conditions; The output end of the data standardization unit is connected to the input end of the regression model unit; the output end of the regression model unit is connected to the input end of the variable screening unit; the output end of the variable screening unit is connected to the input end of the classification prediction module.

[0008] The classification prediction module includes a classification model unit and a classification prediction unit; The classification model unit is used to construct and train an SVM classification model based on the screened independent variables as features and combined with manual annotations; The classification prediction unit is used to standardize the newly collected data and predict the equipment status according to the trained SVM classification model; The output end of the classification model unit is connected to the input end of the classification prediction unit.

[0009] A method for production management of optoelectronic products based on the Internet of Things, the method comprising the following steps: Step 1: Arrange sensors to collect multiple batches of data and manually mark the equipment status corresponding to different batches to obtain the product qualification rate; Step 2: Standardize the acquired data; construct a regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable, and select the independent variable according to the independent variable coefficient of the regression model; Step 3: Using the screened independent variables as features and combining them with manual annotation, a classification model is constructed to standardize the newly collected data and predict the equipment status.

[0010] In step 1, multiple sensors are deployed on the optoelectronic product production equipment to monitor the operating status of the equipment in real time; the sensors are represented as: [C1, C2, …, C n ]; where n is a positive integer, indicating the number of sensors; C1~C n Respectively represent the 1st to nth sensors; Collect multiple batches of equipment data, and the batches are represented as: [B1, B2, …, B u ]; where u is a positive integer, indicating the number of data collection batches; B1~B u Respectively represent the 1st to uth batches; For batch B i , the corresponding data is expressed as: [C1(B i ),C2(B i ),…,C n (B i ),R(B i )]; where i is a positive integer, representing a batch sequence; B i represents the i-th batch, Bi∈{B1,B2,…,B u}; C1(B i )~C n (B i ) represent B i Batch collection to sensors C1~C n data; The equipment status of different batches is annotated manually. The equipment status is expressed as: [K1, K2, …, K a ]; where a is a positive integer, indicating the number of device states, K1~K a Respectively represent 1 to a device status; According to the production report information, the qualified rate of products in different batches is obtained, which is expressed as: [R(B1), R(B2),…, R(B u )]; among them, R(B1)~R(B u ) represent batches B1 to B u The corresponding product qualification rate.

[0011] In step 2, batches B1 to B u The collected data is processed by Z-Score standardization: for sensor C j The collected data is used to calculate C j (B1)~C j (B u) average value C j (mean) and standard deviation C j (std); standardized data is represented by S j (B1)~S j (B u );S j (B1)=(C j (B1)-C j (mean)) / C j (std); where j is a positive integer, representing the sensor sequence; C j represents the jth sensor, C j ∈{C1,C2,…,C n}; S j (B1)~S j (B u ) represent C j (B1)~C j (B u ) The data after standardization; Standardize the data collected by different sensors in different batches; For the data collected from different batches and their corresponding product qualification rates, a polynomial regression model is constructed: ; Among them, R represents the product qualification rate, the dependent variable of the model; S1~S n represents the independent variable, which is the data collected by different sensors after standardization; β0 is a constant term, which represents the expected value of the product qualification rate; β1~β n Represents the coefficient of the first-order term, indicating the corresponding independent variable S1~S n The average change in product qualification rate for each additional unit; β n+1 ~β 2n Represents the quadratic term coefficient, corresponding to the square term S1 of the independent variable 2 ~S n 2 ; m is a positive integer, indicating the number of cross terms, 2n+m <u-1;β 2n+1 ~β 2n+m It represents the cross-term coefficient, which corresponds to the product term between the independent variables and is used to represent the influence of the interaction between different sensor data on the product qualification rate; represents the error term; S' represents the product combination of different sensor data; Combined with the collected data, the least squares method is used to fit the polynomial regression model: the initial value of the coefficient is given, and the gradient descent optimization algorithm is used to continuously adjust the coefficient value, and iterate continuously to ensure that the sum of the squares of the errors between the actual product qualification rate value and the model prediction value meets the preset conditions.

[0012] In step 2, the coefficients obtained after fitting are expressed as β1'~β 2n+m '; Prediction value vector y'=Xβ'; Residual vector ; Covariance matrix Cov(β')=σ 2 (X T X) -1 ; Among them, X represents the independent variable matrix, the size is u×(2n+m), , each row represents a batch, and each column represents the first-order term, second-order term, and cross-term of the standardized data of the batch; β' represents the coefficient matrix after fitting, with a size of (2n+m)×1, β'=[β1',...,β 2n+m '] T ; y represents the product qualification rate vector, y=[R(B1),R(B2),…,R(B u )] T ; σ 2 represents the variance of the residual, σ 2 =( ' T ') / (u-(2n+m)-1); For the coefficient β r ', standard error SE(β r ') = Cov(β') rr 1 / 2, t statistic t r =β r ' / SE(β r '); where r is a positive integer, r∈{1,2,...,2n}; Cov(β') rr Represents the rth element on the diagonal of the covariance matrix Cov(β'); Calculate β1'~β 2n 'T statistic for each coefficient; set the significance level α and get the critical value t according to the t distribution table α / 2 ; For a certain independent variable, when there is a corresponding linear term coefficient or quadratic term coefficient t statistic with an absolute value greater than t α / 2 When the absolute values ​​of the linear term coefficient and the quadratic term coefficient t statistics of an independent variable are both less than or equal to t α / 2 When , it is considered that the independent variable has little effect on the product qualification rate, and the independent variable is removed; the screened independent variable is used to represent the features after dimensionality reduction.

[0013] In step 3, the selected independent variables represent the features after dimensionality reduction, expressed as: [F1, F2, ..., F b]; where b is a positive integer, indicating the number of features after screening, b≤n, F1~F b Represent the 1st to bth features respectively; The SVM classification model is constructed and trained with the standardized sensor data and equipment status corresponding to different batches as input; when new sensor data is collected, it is standardized and then the classification model is used to determine the equipment status corresponding to the batch.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention performs standardized processing on the collected data, so that the data of different sensors in different batches can be compared and analyzed at the same scale, which is conducive to mining the hidden rules in the data; the present invention constructs a polynomial regression model with standardized data as the independent variable and product qualification rate as the dependent variable, and through fitting and coefficient analysis, it can screen out the independent variables that have a significant impact on the product qualification rate; the present invention obtains the reduced-dimensional features by screening the independent variables, and constructs a classification model to predict the equipment status; this feature screening process removes the interference of factors that have a smaller impact on the product qualification rate, so that the classification model can focus more on key factors, thereby improving the accuracy of equipment status prediction for newly collected data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of an optoelectronic product production management system based on the Internet of Things according to the present invention; Figure 2 This is a schematic diagram of the steps of a photoelectric product production management method based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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.

[0017] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a photoelectric product production management system based on the Internet of Things, the system includes a preprocessing module, a feature engineering module and a classification prediction module; The preprocessing module is used to arrange sensors to collect multiple batches of data and manually mark the equipment status corresponding to different batches to obtain the product qualification rate; The feature engineering module is used to standardize the acquired data, and to construct a regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable, and to select the independent variable as the feature according to the independent variable coefficient of the regression model; The classification prediction module is used to construct a classification model based on the screened independent variables as features and combined with manual annotation to standardize the newly collected data and predict the equipment status; The output end of the preprocessing module is connected to the input end of the feature engineering module; the output end of the feature engineering module is connected to the input end of the classification prediction module.

[0018] The preprocessing module includes a data acquisition unit, a manual labeling unit and a product reporting unit; The data acquisition unit is used to arrange sensors on optoelectronic product production equipment and collect multiple batches of data; The manual labeling unit is used to manually label the device status of the collected multiple batches of data; The product reporting unit is used to obtain the production qualification rate of the optoelectronic products according to the production report of each batch of products; The output end of the data acquisition unit is connected to the input end of the manual labeling unit; the output end of the manual labeling unit is connected to the input end of the product reporting unit; the output end of the product reporting unit is connected to the input end of the feature engineering module.

[0019] The feature engineering module includes a data standardization unit, a regression model unit and a variable screening unit; The data standardization unit is used to perform Z-Score standardization processing on the acquired data; The regression model unit is used to construct a polynomial regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable; The variable screening unit is used to calculate the t-statistic of the coefficient of the independent variable of the polynomial regression model and screen the independent variable as a feature according to the preset conditions; The output end of the data standardization unit is connected to the input end of the regression model unit; the output end of the regression model unit is connected to the input end of the variable screening unit; the output end of the variable screening unit is connected to the input end of the classification prediction module.

[0020] The classification prediction module includes a classification model unit and a classification prediction unit; The classification model unit is used to construct and train an SVM classification model based on the screened independent variables as features and combined with manual annotations; The classification prediction unit is used to standardize the newly collected data and predict the equipment status according to the trained SVM classification model; The output end of the classification model unit is connected to the input end of the classification prediction unit.

[0021] A method for production management of optoelectronic products based on the Internet of Things, the method comprising the following steps: Step 1: Arrange sensors to collect multiple batches of data and manually mark the equipment status corresponding to different batches to obtain the product qualification rate; Step 2: Standardize the acquired data; construct a regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable, and select the independent variable according to the independent variable coefficient of the regression model; Step 3: Using the screened independent variables as features and combining them with manual annotation, a classification model is constructed to standardize the newly collected data and predict the equipment status.

[0022] In step 1, multiple sensors are deployed on the optoelectronic product production equipment to monitor the operating status of the equipment in real time; the sensors are represented as: [C1, C2, …, C n ]; where n is a positive integer, indicating the number of sensors; C1~C n Represent the 1st to nth sensors respectively; Collect multiple batches of equipment data, and the batches are represented as: [B1, B2, …, B u ]; where u is a positive integer, indicating the number of data collection batches; B1~B u Represent the 1st to uth batches respectively; For batch B i , the corresponding data is expressed as: [C1(B i ),C2(B i ),…,C n (B i ),R(B i )]; where i is a positive integer, representing a batch sequence; B i represents the i-th batch, Bi∈{B1,B2,…,B u}; C1(B i )~C n (B i ) represent B i Batch collection to sensors C1~C n data; The equipment status of different batches is annotated manually. The equipment status is represented as: [K1, K2, …, K a ]; where a is a positive integer, indicating the number of device states, K1~K a Respectively represent 1 to a device status; According to the production report information, the qualified rate of products in different batches is obtained, which is expressed as: [R(B1), R(B2),…, R(B u )]; among them, R(B1)~R(B u ) represent batches B1 to Bu The corresponding product qualification rate.

[0023] In step 2, batches B1 to B u The collected data is processed by Z-Score standardization: for sensor C j The collected data is used to calculate C j (B1)~C j (B u ) average value C j (mean) and standard deviation C j (std); standardized data is represented by S j (B1)~S j (B u );S j (B1)=(C j (B1)-C j (mean)) / C j (std); where j is a positive integer, representing the sensor sequence; C j represents the jth sensor, C j ∈{C1,C2,…,C n}; S j (B1)~S j (B u ) represent C j (B1)~C j (B u ) The data after standardization; Standardize the data collected by different sensors in different batches; For the data collected from different batches and their corresponding product qualification rates, a polynomial regression model is constructed: ; Among them, R represents the product qualification rate, the dependent variable of the model; S1~S n represents the independent variable, which is the data collected by different sensors after standardization; β0 is a constant term, which represents the expected value of the product qualification rate; β1~β n Represents the coefficient of the first-order term, indicating the corresponding independent variable S1~S n The average change in product qualification rate for each additional unit; β n+1 ~β 2n Represents the quadratic term coefficient, corresponding to the square term S1 of the independent variable 2 ~S n 2 ; m is a positive integer, indicating the number of cross terms, 2n+m <u-1;β 2n+1 ~β 2n+mIt represents the cross-term coefficient, which corresponds to the product term between the independent variables and is used to represent the influence of the interaction between different sensor data on the product qualification rate; represents the error term; S' represents the product combination of different sensor data; Combined with the collected data, the least squares method is used to fit the polynomial regression model: the initial value of the coefficient is given, and the gradient descent optimization algorithm is used to continuously adjust the coefficient value, and iterate continuously to ensure that the sum of the squares of the errors between the actual product qualification rate value and the model prediction value meets the preset conditions.

[0024] In step 2, the coefficients obtained after fitting are expressed as β1'~β 2n+m '; Prediction value vector y'=Xβ'; Residual vector ; Covariance matrix Cov(β')=σ 2 (X T X) -1 ; Among them, X represents the independent variable matrix, the size is u×(2n+m), , each row represents a batch, and each column represents the first-order term, second-order term, and cross-term of the standardized data of the batch; β' represents the coefficient matrix after fitting, with a size of (2n+m)×1, β'=[β1',...,β 2n+m '] T ; y represents the product qualification rate vector, y=[R(B1),R(B2),…,R(B u )] T ; σ 2 represents the variance of the residual, σ 2 =( ' T ') / (u-(2n+m)-1); For the coefficient β r ', standard error SE(β r ') = Cov(β') rr 1 / 2, t statistic t r =β r ' / SE(β r '); where r is a positive integer, r∈{1,2,...,2n}; Cov(β') rr Represents the rth element on the diagonal of the covariance matrix Cov(β'); Calculate β1'~β 2n 'T statistic for each coefficient; set the significance level α and get the critical value t according to the t distribution table α / 2 ; For a certain independent variable, when there is a corresponding linear term coefficient or quadratic term coefficient t statistic with an absolute value greater than t α / 2When the absolute values ​​of the linear term coefficient and the quadratic term coefficient t statistics of an independent variable are both less than or equal to t α / 2 When , it is considered that the independent variable has little effect on the product qualification rate, and the independent variable is removed; the screened independent variable is used to represent the features after dimensionality reduction.

[0025] In step 3, the selected independent variables represent the features after dimensionality reduction, expressed as: [F1, F2, ..., F b ]; where b is a positive integer, indicating the number of features after screening, b≤n, F1~F b Represent the 1st to bth features respectively; The SVM classification model is constructed and trained with the standardized sensor data and equipment status corresponding to different batches as input; when new sensor data is collected, it is standardized and then the classification model is used to determine the equipment status corresponding to the batch.

[0026] In this embodiment, a certain optoelectronic product manufacturer produces a precision photodetector, and its production process is complex, involving multiple key production equipment, including photolithography equipment, coating equipment, packaging equipment, etc.; Step 1: Arrange sensors to collect multiple batches of data and manually mark the equipment status corresponding to different batches to obtain the product qualification rate; Five sensors C1-C5 are deployed on the lithography equipment to monitor the exposure time C1, photoresist thickness C2, light source intensity C3, workbench temperature C4 and alignment accuracy C5 of lithography; four sensors C6-C9 are deployed on the coating equipment to monitor the coating rate C6, film thickness C7, vacuum degree C8 and evaporation source temperature C9; three sensors C10-C12 are deployed on the packaging equipment to monitor the packaging pressure C10, glue dosage C11 and curing time C12; A total of 20 batches of data from B1 to B20 were collected. For batch B5, the data collected by the lithography equipment were [C1(B5)=15, C2(B5)=0.8, C3(B5)=1200, C4(B5)=25, C5(B5)=0.05], and the data collected by the coating equipment were [C6(B5)=0.5, C7(B5)=100, C8(B5)=5×10 -4 ,C9(B5)=1800], and the data collected by the packaging equipment is [C10(B5)=0.8,C11(B5)=0.3,C12(B5)=60].

[0027] Through manual inspection and analysis of production process records, the equipment status of each batch is marked. Assume that the equipment status is divided into three states: normal K1, minor fault K2 and serious fault K3; the equipment status of batch B5 is manually marked as minor fault K2; According to the final product inspection report, the qualified rate of each batch of products is obtained; the qualified rate of batch B5 R(B5)=85%.

[0028] Step 2: Standardize the acquired data; construct a regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable, and select the independent variable according to the independent variable coefficient of the regression model; Perform Z-Score standardization on the 20 batches of data collected; take sensor C1 as an example, calculate its mean value C1(mean) and standard deviation C1(std) in the 20 batches of data, and then standardize the data of each batch; for the C1(B5) data of batch B5, after standardization, S1(B5)=(15-C1(mean)) / C1(std) is obtained; Construct a polynomial regression model, with the dependent variable being the product qualification rate R and the independent variable being the standardized sensor data; the constructed model is: R = β0 + β1S1 + β2S2 + ⋯ + β 12 S 12 +β 13 S1 2 +β 14 S2 2 +⋯+β 24 S 12 2 +β 25 S1S2+⋯+β 35 S'+ϵ; The least squares method is used to fit the polynomial regression model. The initial value of the coefficient is given, and the gradient descent optimization algorithm is used to continuously adjust the coefficient value. After multiple iterations, the fitted coefficients β1'~β 35 '; Calculate the t statistic for each coefficient, assuming that the significance level α=0.05 is set, and the critical value t is obtained according to the t distribution table α / 2 ; After calculation and comparison, it is found that the absolute value of the linear coefficient or quadratic coefficient t statistic corresponding to the independent variables S1 exposure time, S3 light source intensity, S7 film thickness, S8 vacuum degree and S10 packaging pressure is greater than t α / 2 , it is believed that these independent variables have a significant impact on the product qualification rate, and these independent variables are retained as the features after screening.

[0029] Step 3: Using the screened independent variables as features and combining them with manual annotation, a classification model is constructed to standardize the newly collected data and predict the equipment status; The screened independent variables are used as the features after dimensionality reduction, i.e. [F1=S1, F2=S3, F3=S7, F4=S8, F5=S10]; The SVM classification model is constructed and trained using the standardized sensor data and equipment status corresponding to different batches as input; the 20 batches of data collected previously are divided into 15 training batches and 5 test batches, and the SVM classification model is trained using the training set, and the parameters of the model are adjusted so that it can accurately classify the equipment status; When new sensor data is collected, the new batch B21, its sensor data is standardized to obtain the standardized feature values ​​[F1(B21), F2(B21), F3(B21), F4(B21), F5(B21)], and then these feature values ​​are input into the trained SVM classification model to determine the device status corresponding to the batch; The model predicts that the equipment status of batch B21 is normal K1. Production managers can arrange production tasks and equipment maintenance plans reasonably based on this prediction result to ensure the smooth progress of the production process and the stability of product quality.

[0030] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An optoelectronic product production management system based on the Internet of Things, characterized by: The system includes a preprocessing module, a feature engineering module, and a classification prediction module; The preprocessing module is used to arrange sensors to collect multiple batches of data and manually mark the equipment status corresponding to different batches to obtain the product qualification rate; The feature engineering module is used to standardize the acquired data, and to construct a regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable, and to select the independent variable as the feature according to the independent variable coefficient of the regression model; The classification prediction module is used to construct a classification model based on the screened independent variables as features and combined with manual annotation to standardize the newly collected data and predict the equipment status; The output end of the preprocessing module is connected to the input end of the feature engineering module; the output end of the feature engineering module is connected to the input end of the classification prediction module.

2. According to the optoelectronic product production management system based on the Internet of Things according to claim 1, it is characterized by: The preprocessing module includes a data acquisition unit, a manual labeling unit and a product reporting unit; The data acquisition unit is used to arrange sensors on optoelectronic product production equipment and collect multiple batches of data; The manual labeling unit is used to manually label the device status of the collected multiple batches of data; The product reporting unit is used to obtain the production qualification rate of the optoelectronic products according to the production report of each batch of products; The output end of the data acquisition unit is connected to the input end of the manual labeling unit; the output end of the manual labeling unit is connected to the input end of the product reporting unit; the output end of the product reporting unit is connected to the input end of the feature engineering module.

3. According to claim 2, the optoelectronic product production management system based on the Internet of Things is characterized by: The feature engineering module includes a data standardization unit, a regression model unit and a variable screening unit; The data standardization unit is used to perform Z-Score standardization processing on the acquired data; The regression model unit is used to construct a polynomial regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable; The variable screening unit is used to calculate the t-statistic of the coefficient of the independent variable of the polynomial regression model and screen the independent variable as a feature according to the preset conditions; The output end of the data standardization unit is connected to the input end of the regression model unit; the output end of the regression model unit is connected to the input end of the variable screening unit; the output end of the variable screening unit is connected to the input end of the classification prediction module.

4. The optoelectronic product production management system based on the Internet of Things according to claim 3 is characterized by: The classification prediction module includes a classification model unit and a classification prediction unit; The classification model unit is used to construct and train an SVM classification model based on the screened independent variables as features and combined with manual annotations; The classification prediction unit is used to standardize the newly collected data and predict the equipment status according to the trained SVM classification model; The output end of the classification model unit is connected to the input end of the classification prediction unit.

5. A method for production management of optoelectronic products based on the Internet of Things, characterized in that: The method comprises the following steps: Step 1: Arrange sensors to collect multiple batches of data and manually mark the equipment status corresponding to different batches to obtain the product qualification rate; Step 2: Standardize the acquired data; construct a regression model with the standardized data as the independent variable and the product qualification rate as the dependent variable, and select the independent variable according to the independent variable coefficient of the regression model; Step 3: Using the screened independent variables as features and combining them with manual annotation, a classification model is constructed to standardize the newly collected data and predict the equipment status.

6. The method for managing optoelectronic product production based on the Internet of Things according to claim 5, characterized in that: In step 1, multiple sensors are deployed on the optoelectronic product production equipment to monitor the operating status of the equipment in real time; the sensors are represented as: [C1, C2, …, C n ]; where n is a positive integer, indicating the number of sensors; C1~C n Represent the 1st to nth sensors respectively; Collect multiple batches of equipment data, and the batches are represented as: [B1, B2, …, B u ]; where u is a positive integer, indicating the number of data collection batches; B1~B u Respectively represent the 1st to uth batches; According to the production report information, the qualified rate of products in different batches is obtained, which is expressed as: [R(B1), R(B2),…, R(B u )]; among them, R(B1)~R(B u ) represent batches B1 to B u Corresponding product qualification rate; For batch B i , the corresponding data is expressed as: [C1(B i ),C2(B i ),…,C n (B i ),R(B i )]; where i is a positive integer, representing a batch sequence; B i represents the i-th batch, Bi∈{B1,B2,…,B u }; C1(B i )~C n (B i ) represent B i Batch collection to sensors C1~C n data; The equipment status of different batches is annotated manually. The equipment status is represented as: [K1, K2, …, K a ]; where a is a positive integer, indicating the number of device states, K1~K a Respectively represent 1 to a device states.

7. The method for managing optoelectronic product production based on the Internet of Things according to claim 6, characterized in that: In step 2, batches B1 to B u The collected data is processed by Z-Score standardization: for sensor C j The collected data is used to calculate C j (B1)~C j (B u ) average value C j (mean) and standard deviation C j (std); standardized data is represented by S j (B1)~S j (B u );S j (B1)=(C j (B1)-C j (mean)) / C j (std); where j is a positive integer, representing the sensor sequence; C j represents the jth sensor, C j ∈{C1,C2,…,C n }; S j (B1)~S j (B u ) represent C j (B1)~C j (B u ) The data after standardization; Standardize the data collected by different sensors in different batches; For the data collected from different batches and their corresponding product qualification rates, a polynomial regression model is constructed: ； Among them, R represents the product qualification rate, the dependent variable of the model; S1~S n represents the independent variable, which is the data collected by different sensors after standardization; β0 is a constant term, which represents the expected value of the product qualification rate; β1~β n Represents the coefficient of the first-order term, indicating the corresponding independent variable S1~S n The average change in product qualification rate for each additional unit; β n+1 ~β 2n Represents the quadratic term coefficient, corresponding to the square term S1 of the independent variable 2 ~S n 2 ; m is a positive integer, indicating the number of cross terms, 2n+m <u-1;β 2n+1 ~β 2n+m It represents the cross-term coefficient, which corresponds to the product term between the independent variables and is used to represent the influence of the interaction between different sensor data on the product qualification rate; represents the error term; S' represents the product combination of different sensor data; Combined with the collected data, the least squares method is used to fit the polynomial regression model: the initial value of the coefficient is given, and the gradient descent optimization algorithm is used to continuously adjust the coefficient value, and iterate continuously to ensure that the sum of the squares of the errors between the actual product qualification rate value and the model prediction value meets the preset conditions.

8. The method for managing optoelectronic product production based on the Internet of Things according to claim 7, characterized in that: In step 2, the coefficients obtained after fitting are expressed as β1'~β 2n+m '; Prediction value vector y'=Xβ'; Residual vector ; Covariance matrix Cov(β')=σ 2 (X T X) -1 ; Among them, X represents the independent variable matrix, the size is u×(2n+m), , each row represents a batch, and each column represents the first-order term, second-order term, and cross-term of the standardized data of the batch; β' represents the coefficient matrix after fitting, with a size of (2n+m)×1, β'=[β1',...,β 2n+m '] T ; y represents the product qualification rate vector, y=[R(B1),R(B2),…,R(B u )] T ; σ 2 represents the variance of the residual, σ 2 =( ' T ') / (u-(2n+m)-1); For the coefficient β r ', standard error SE (β r ') = Cov(β') rr 1 / 2, t statistic t r =β r ' / SE(β r '); where r is a positive integer, r∈{1,2,...,2n}; Cov(β') rr Represents the rth element on the diagonal of the covariance matrix Cov(β'); Calculate β1'~β 2n 'T statistic for each coefficient; set the significance level α and get the critical value t according to the t distribution table α / 2 ; For a certain independent variable, when there is a corresponding linear term coefficient or quadratic term coefficient t statistic with an absolute value greater than t α / 2 When the absolute values ​​of the linear term coefficient and the quadratic term coefficient t statistics of an independent variable are both less than or equal to t α / 2 When , it is considered that the independent variable has little effect on the product qualification rate, and the independent variable is removed; the screened independent variable is used to represent the features after dimensionality reduction.

9. The method for managing optoelectronic product production based on the Internet of Things according to claim 8, characterized in that: In step 3, the selected independent variables represent the features after dimensionality reduction, expressed as: [F1, F2, ..., F b ]; where b is a positive integer, indicating the number of features after screening, b≤n, F1~F b Represent the 1st to bth features respectively; The SVM classification model is constructed and trained with the standardized sensor data and equipment status corresponding to different batches as input; when new sensor data is collected, it is standardized and then the classification model is used to determine the equipment status corresponding to the batch.

Images