A method for identifying key production factors for multi-stage product manufacturing

Abstract

The application provides a key production factor identification method for multi-stage product manufacturing, and relates to the technical field of industrial process design. Firstly, continuous production variables and discrete production variables of each production stage are collected in real time along the production line in the multi-stage product production line, the detection time of the product at the final inspection position is taken as a global time anchor point, the time lag of the product from each production stage to the final inspection position is accurately calculated, the continuous and discrete production variables are integrated based on the time lag, and the continuous production variables are subjected to feature processing, so that a production factor data set corresponding to a specific product is formed, a hierarchical sparse partial function linear regression model is constructed and trained, the integrated production factor data set is input into the model, and key production factors are output. The application realizes accurate alignment of multi-modal production data and significantly improves the accuracy and efficiency of identification of key production stages, production processes and production variables affecting product quality.

Description

Technical Field

[0001] This invention relates to the technical field of industrial process design, and more specifically, to a method for identifying key production factors for multi-stage product manufacturing. Background Technology

[0002] In product manufacturing processes involving multiple steps, products typically undergo several production stages from raw material handling to final inspection. To monitor production status in real time, various types of sensors, such as pressure sensors, fan frequency sensors, and temperature sensors, are usually deployed along the production line. Based on these sensors, the production line can collect a large amount of multimodal data during the production process to assist in adjusting production process parameters. This multimodal data includes functional data such as temperature, as well as scalar data such as pressure and fan frequency.

[0003] However, the quality of the final inspected product is affected by variables from multiple processes in a complex multi-stage production process. There is a significant time lag between the data collected by the sensors and the final inspection data. This lag varies with the production speed. Some sensors also cover multiple production stages, making it difficult to accurately determine the start and end times of the data window. This makes it difficult to achieve unified alignment of multi-source data at the product level. Therefore, it is difficult to identify the key variables affecting the quality of the final product by directly analyzing the collected data. When a batch of products has quality defects, due to the misalignment of multi-source data, it is impossible to trace back to the specific process, equipment, or parameter abnormality, which directly affects the accuracy of quality control and the feasibility of production optimization.

[0004] To identify key production variables, the first step is to address the difficulty in unifying and aligning multimodal data in the production process. Existing technologies synchronize multimodal data from each production stage by setting a fixed time lag and then unifying the multimodal data into scalars for fusion. While this approach achieves unified alignment of multimodal data in the production process to some extent, it also loses key data information and struggles to adapt to dynamic fluctuations in production speed. Summary of the Invention

[0005] To address the problems of existing methods for identifying key production factors, such as difficulty in aligning and unifying multimodal production data and the easy loss of key production data, this invention proposes a key production factor identification method for multi-stage product manufacturing. By dynamically calculating lag time and using principal component analysis, the method achieves unified alignment of multimodal production data and preserves key production data in the production process.

[0006] To achieve the above-mentioned technical effects, the technical solution of the present invention is as follows: Firstly, this application proposes a method for identifying key production factors in multi-stage product manufacturing, wherein the key production factors include: production stage, production process, and production variables; the method includes the following steps: In a multi-stage product production line, production variables for each production stage are collected in real time along the production line. These production variables include continuous production variables and discrete production variables. Using the inspection time of the product at the final inspection position as the global time anchor point, and based on the conveying speed of the production line, the time lag of the product from each production stage to the final inspection position is calculated in multi-stage product production. Based on time lag, continuous production variables and discrete production variables are integrated separately, and the continuous production variables are processed to form a production factor dataset corresponding to the product. A hierarchical sparse partial function linear regression model is constructed and trained. The integrated production factor dataset is then input into the trained hierarchical sparse partial function linear regression model to obtain the key production factors.

[0007] In this technical solution, continuous and discrete production variables are first collected in real time along the multi-stage product production line for each production stage. Using the inspection time of the product at the final inspection position as the global time anchor, the time lag of the product's flow from each production stage to the final inspection position is accurately calculated. Based on this time lag, the continuous and discrete production variables are integrated, and feature processing is performed on the continuous production variables to form a production factor dataset corresponding to a specific product. A hierarchical sparse partial function linear regression model is constructed and trained. The integrated production factor dataset is input into this model, and key production factors are output. This achieves accurate alignment of multimodal production data and significantly improves the accuracy and efficiency of identifying key production stages, processes, and variables affecting product quality.

[0008] Preferably, when collecting production variables at each production stage in real time along the production line, sensors deployed at each production stage are used to collect the production variables of the product at each production stage. The process of calculating the time lag from each production stage to the final inspection position is as follows: Based on the inspection time of the product at the final inspection location Construct global time anchors; Obtain the physical transmission distance between each production stage and the final inspection location. , Indicates the production stage sequence, based on the physical transmission distance. Conveyor speed of the production line in the corresponding production stage Calculate the time lag of the product at each stage of production. ; Based on the time lag of products at different stages of production and global time anchor Calculate the time lag of the product at each stage of production on the production line; the time lag includes the instantaneous lag of discrete production variables. Lag at both ends of continuous production variables .

[0009] Preferably, the integration of continuous production variables and discrete production variables based on time lag includes: Based on the instantaneous lag of the discrete production variables Lag at both ends of continuous production variables Obtain products At each production stage, observations are conducted, and fragments of discrete and continuous production variables are obtained for each stage.

[0010] Preferably, the processing of continuous production variables includes: extracting the continuous production variables with lags at both ends based on principal component analysis. Segments within the interval Principal component scores; The production factor dataset corresponding to the product includes: by product Summary of continuous production variables is lagging at both ends The principal component scores and discrete production variables within the interval segments form a production factor dataset.

[0011] Preferably, the hierarchical sparse partial function linear regression model includes an objective function and constraints, wherein the expression for the objective function is:

[0012] The expression for the constraint condition is:

[0013]

[0014] in, This represents the weighting coefficient for the production stage. Indicates the weighting coefficient of the production process. This represents the weighting coefficients of the production variable. The regularization coefficient represents the weighting factor of the production stage. The regularization coefficient represents the weighting factor of the production process. The regularization coefficient represents the weighting coefficient of the production variable. Indicates the production stage group. Indicates the group of the production process.

[0015] Preferably, before training the hierarchical sparse partial function linear regression model, the method further includes: automatically selecting the regularization coefficients of the production stage weight coefficients using the Bayesian information criterion. Regularization coefficient of production process weighting coefficient Regularization coefficients of production variable weights .

[0016] Preferably, during the training of the hierarchical sparse partial function linear regression model, the production stage weight coefficients in the model are updated alternately using block coordinate descent. Production process weighting coefficient and production variable weighting coefficients .

[0017] Secondly, this application also proposes a key variable factor identification system for multi-stage product manufacturing, the system comprising: The production variable acquisition unit is used to collect production variables in real time along the production line of a multi-stage product production line. The production variables include continuous production variables and discrete production variables. The time lag calculation unit is used to calculate the time lag of the product from each production stage to the final inspection position in multi-stage product production, based on the conveyor speed of the production line, using the inspection time of the product at the final inspection position as the global time anchor point. The production factor dataset construction unit is used to integrate continuous and discrete production variables based on time lag, and process the continuous production variables to form a production factor dataset corresponding to the product. The key production factor solution unit is used to construct and train a hierarchical sparse partial function linear regression model. The integrated production factor dataset is input into the trained hierarchical sparse partial function linear regression model and solved to obtain the key production factors.

[0018] Thirdly, this application also proposes a computer device, which includes a memory, a processor, and a computer program stored in the memory that can be run on the processor. The processor executes the computer program to implement the aforementioned method for identifying key production factors for multi-stage product manufacturing.

[0019] Fourthly, this application also proposes a computer storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to execute the aforementioned method for identifying key production factors for multi-stage product manufacturing.

[0020] Compared with the prior art, the beneficial effects of the present invention are: This invention proposes a method for identifying key production factors in multi-stage product manufacturing. First, continuous and discrete production variables are collected in real-time along the multi-stage production line for each stage. Using the inspection time of the product at the final inspection position as a global time anchor, the time lag of the product's flow from each production stage to the final inspection position is accurately calculated. Based on this time lag, the continuous and discrete production variables are integrated, and feature processing is performed on the continuous production variables to form a production factor dataset corresponding to a specific product. A hierarchical sparse partial function linear regression model is constructed and trained. The integrated production factor dataset is input into this model, and the key production factors are output. This method achieves accurate alignment of multimodal production data and significantly improves the accuracy and efficiency of identifying key production stages, processes, and variables affecting product quality. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating a method for identifying key production factors in multi-stage product manufacturing, as proposed in Embodiment 1 of the present invention. Figure 2 This is a schematic diagram illustrating the process of integrating production variables in a multi-stage product manufacturing process as proposed in Embodiment 2 of the present invention. Figure 3 This is a schematic diagram of the structure of a key production factor identification system for multi-stage product manufacturing proposed in Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the structure of the computer device proposed in Embodiment 2 of the present invention. Detailed Implementation

[0022] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some parts of the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions; It is understandable to those skilled in the art that some well-known details may be omitted from the accompanying drawings.

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] The positional relationships depicted in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent. Example 1 This embodiment proposes a method for identifying key production factors in multi-stage product manufacturing. The key production factors include: production stage, production process, and production variables. A flowchart illustrating this method can be found here. Figure 1 This includes the following steps: S1. In a multi-stage product production line, production variables for each production stage are collected in real time along the production line. The production variables include continuous production variables and discrete production variables. S2. Using the inspection time of the product at the final inspection position as the global time anchor point, and based on the conveying speed of the production line, calculate the time lag of the product from each production stage to the final inspection position in multi-stage product production. S3. Based on time lag, integrate continuous production variables and discrete production variables respectively, and process the continuous production variables to form a production factor dataset corresponding to the product; S4. Construct and train a hierarchical sparse partial function linear regression model, and input the integrated production factor dataset into the trained hierarchical sparse partial function linear regression model to obtain the key production factors.

[0025] Specifically, taking the production of ceramic products as an example, a ceramic product production line includes several orderly production stages, such as a preheating stage, a firing stage, and a rapid cooling stage. Each production stage corresponds to one or more production processes. For example, in the preheating stage, the ceramic product may first undergo a drying and dehydration process before going through a heating and preheating process. Among these, the production variables in the drying and dehydration process include the frequency of the exhaust fan. and the pressure of the exhaust duct During the production process, these data only need to be read at a certain point in time. Therefore, the fan frequency and pressure are discrete production variables. The production variables of the heating and preheating process include continuous temperature changes. During the production process, these data need to be read at a certain time period. Therefore, the temperature changes are continuous production variables.

[0026] Specifically, the production factor dataset includes specific production stages, production processes, and production variables; for example, the fan frequency of the dehumidification fan during the drying and dehydration process of the product in the preheating section. This data can be used as one piece of data in a dataset that includes the production stage (i.e., the preheating stage), the production process (i.e., the drying and dehydration process), and production variables (i.e., the fan frequency). ).

[0027] In this embodiment, continuous and discrete production variables are first collected in real time along the multi-stage product production line for each production stage. Using the inspection time of the product at the final inspection position as the global time anchor, the time lag of the product's flow from each production stage to the final inspection position is accurately calculated. Based on this time lag, the continuous and discrete production variables are integrated, and feature processing is performed on the continuous production variables to form a production factor dataset corresponding to a specific product. A hierarchical sparse partial function linear regression model is constructed and trained. The integrated production factor dataset is input into this model, and key production factors are output. This achieves accurate alignment of multimodal production data and significantly improves the accuracy and efficiency of identifying key production stages, processes, and variables affecting product quality.

[0028] Example 2 In this embodiment, when collecting production variables at each production stage in real time along the production line, sensors deployed at each production stage are used to collect the production variables of the product at each production stage. The process of calculating the time lag from each production stage to the final inspection position is as follows: Based on the inspection time of the product at the final inspection location Construct global time anchors; Obtain the physical transmission distance between each production stage and the final inspection location. , Indicates the production stage sequence, based on the physical transmission distance. Conveyor speed of the production line in the corresponding production stage Calculate the time lag of the product at each stage of production. ; Based on the time lag of products at different stages of production and global time anchor Calculate the time lag of the product at each stage of production on the production line; the time lag includes the instantaneous lag of discrete production variables. Lag at both ends of continuous production variables .

[0029] Specifically, taking the production of ceramic products as an example, when collecting production variables at each stage of the ceramic production line in real time, the production line is first divided into sections based on the actual physical process flow. The production process is divided into several orderly stages, such as the preheating stage, the firing stage, and the quenching stage.

[0030] Specifically, at each stage of production The sensors deployed include a first type of sensor and a second type of sensor. The first type of sensor, such as a pressure sensor or a fan frequency sensor, is used to collect pressure or fan frequency variables that have only a single value at a given moment. The second type of sensor, such as a thermocouple or an infrared thermometer, is used to continuously record the temperature changes in the area along the time axis, forming a continuous production variable.

[0031] Specifically, considering the potential for significant errors or high hardware costs in directly measuring conveyor belt speed at industrial production sites, the operating frequency of the conveyor motor in this embodiment is... Calculate the conveyor speed of the real-time production line. The expression is:

[0032] in, This represents the preset speed conversion coefficient, which is obtained through Faraday's law and the mechanical parameters of the transmission device.

[0033] Specifically, the time lag of the product in each stage of production on the production line is mainly calculated using a global anchor point reverse backtracking method, the process of which is as follows: When product When the product reaches the quality inspection station at the end of the production line, record the time at that moment. At that moment Defined as a product The unique global time anchor point, and all historical data moments of this product are calculated backward from this point; Get the Physical transmission distance of each production stage Due to the speed of the product when it travels this distance It is variable, and the system can calculate the time lag of the product through this stage using a discrete accumulation method. The expression is:

[0034] In this embodiment, continuous production variables and discrete production variables are integrated based on time lag, including: Based on the instantaneous lag of the discrete production variables Lag at both ends of continuous production variables Obtain products At each production stage, observations are conducted, and fragments of discrete and continuous production variables are obtained for each stage.

[0035] Specifically, before obtaining the discrete production variables in each production stage, it is necessary to find the product... At the instant of the sensor's movement... Specifically, before obtaining the discrete production variables in each production stage, it is necessary to find the product... At the instantaneous point in time of the sensor, first determine the production stage in which the target sensor is located, and then determine all stages after that stage (from...). To the final stage The lag time and the transmission time of the sensor from its own location to the stage exit. The instantaneous lag is obtained by accumulation, and the expression is:

[0036] Based on instantaneous lag and global time anchors, obtain product The timing of collecting discrete production variables at each production stage The expression is:

[0037] Query at the time of collection The sensor readings are used as the product's... Discrete production variables, including pressure or frequency readings.

[0038] Specifically, for continuous production variables, it is necessary to extract product data. The complete curve from entering to leaving the continuous production variable acquisition region yields a segment of the continuous production variable. In this embodiment, temperature is used as a continuous production variable to determine the time point when the product enters the temperature acquisition zone. and the time point of leaving the temperature collection area The expression is:

[0039]

[0040] In the temperature change curve collected by the sensor, cut out from arrive The temperature curve within this time window yields a fragment of the continuous production variable. .

[0041] Specifically, the flowchart illustrating the integration of production variables in the multi-stage product manufacturing process is as follows: Figure 2 As shown, in Figure 2 In this process, the real-time speed of the conveyor belt is calculated by using the frequency of the conveyor motor, the length of the conveyor belt, and the speed change coefficient as input parameters, and then the dynamic transmission time lag of the product in different production stages is calculated. Figure 2 Products were also showcased. The manufacturing process, in the production line, has key production factors including production stages, production processes, and production variables. A production stage comprises different production processes. Specifically, taking the firing stage of ceramic products as an example, its production process includes heating and ventilation processes. Each production process includes one or more production variables; for example, the heating process includes temperature variables, and the ventilation process includes fan frequency variables and pipeline gas pressure variables. These variables are collected by sensors. Based on the calculated time lag and global time anchor point, [the following is a process description]... Figure 2 The data flow frame shown includes continuous sensor data such as pressure change curves and fan frequency fluctuation curves, which are used to trace back and accurately locate the product. Actual process The pressure sensor and the first The specific time of each wind turbine frequency sensor reading (i.e., the positions shown by points A and B on the curve in the figure). Through this backtracking mechanism based on dynamic time delay, data relevant to that specific product can be extracted from continuous time-series data. One-to-one correspondence of instantaneous pressure data and fan frequency data Ultimately, it will come from the entire production line. A pressure sensor and The scattered data from individual wind turbine frequency sensors are integrated into a unified product-level process dataset. and This solves the problem of data asynchrony caused by fluctuations in production speed, and provides accurate input features for the subsequent hierarchical sparse partial function linear regression model.

[0042] In this embodiment, the processing of continuous production variables includes: extracting the continuous production variables with lags at both ends based on principal component analysis. Segments within the interval Principal component scores; Create a production factor dataset corresponding to each product, including: by product Summary of continuous production variables is lagging at both ends The principal component scores and discrete production variables within the interval segments form a production factor dataset.

[0043] Specifically, the principal component analysis process is as follows: first, extract the continuous production variables that are lagged at both ends. Segments within the interval In this embodiment, the segment represents a temperature curve. Next, principal component analysis is performed on these functional objects. By performing eigenvalue decomposition on the covariance function, a set of orthogonal feature waveforms is obtained. Finally, based on the standard that the cumulative variance contribution rate reaches a preset threshold (e.g., 90%), the top few principal components are selected, and the projection weights of the curve on these basis functions are calculated, i.e., the principal component scores. These principal component scores typically represent the overall mean shift of the curve (e.g., overall temperature being too high or too low), the linear trend change of the curve (e.g., rapid or slow heating rate), local fluctuations of the curve, or more complex nonlinear bending patterns. The data ultimately extracted from the temperature curve are these principal component scores, which transform the originally high-dimensional and complex waveform information into a set of concise numerical features, used subsequently to form a production factor dataset with discrete production variables.

[0044] In this embodiment, the hierarchical sparse partial function linear regression model includes an objective function and constraints. The expression for the objective function is as follows:

[0045] The expression for the constraint condition is:

[0046]

[0047] in, This represents the weighting coefficient for the production stage. Indicates the weighting coefficient of the production process. This represents the weighting coefficients of the production variable. The regularization coefficient represents the weighting factor of the production stage. The regularization coefficient represents the weighting factor of the production process. The regularization coefficient represents the weighting coefficient of the production variable. Indicates the production stage group. Indicates the group of the production process.

[0048] Specifically, the objective function is achieved by reparameterizing the regression coefficients to... And on Applying regularization constraints ensures that variable selection must follow physical hierarchy logic. If the optimization process determines that a certain production stage is unimportant (i.e., ... (compressed to 0), the constraint condition guarantees the weight coefficients of all production processes under this stage. and production variable weighting coefficients Automatically force zeroing.

[0049] In this embodiment, before training the hierarchical sparse partial function linear regression model, the method further includes: automatically selecting the regularization coefficient of the production stage weight coefficients using the Bayesian information criterion. Regularization coefficient of production process weighting coefficient Regularization coefficients of production variable weights .

[0050] In this embodiment, during the training of the hierarchical sparse partial function linear regression model, the production stage weight coefficients in the model are updated alternately using block coordinate descent. Production process weighting coefficient and production variable weighting coefficients .

[0051] Specifically, the process of alternately updating the weight coefficients in the model using block coordinate descent is as follows: S41. Initialize production stage weight coefficients Production process weighting coefficient and production variable weighting coefficients ; S42. Weighting coefficients for the production process and production variable weighting coefficients The weight coefficients of the production stage are fixed as constants and the optimal production stage weights are calculated using coordinate descent in the current iteration. S43. Weighting coefficients for production stages The production variable weight coefficients are fixed to the current optimal production stage weight coefficients. The weight coefficients of the production process are fixed as constants, and the optimal production process weights for the current iteration are calculated using coordinate descent. S44. Weighting coefficients for each production stage. and production process weighting coefficient The optimal production stage weight coefficient and the optimal production process weight coefficient under the current iteration are fixed, and the optimal production variable weight coefficient under the current iteration is calculated using coordinate descent. S45. Calculate the objective function value under the current iteration using the optimal production stage weight coefficient, production process weight coefficient, and production variable weight coefficient under the current iteration; S46. Determine if the objective function value is less than a preset threshold. If it is, complete the update, using the current production stage weight coefficient, production process weight coefficient, and production variable weight coefficient as the optimal model parameters to complete the training of the hierarchical sparse partial function linear regression model. If the objective function value is greater than or equal to the preset threshold, return to iterative execution of S42~S45 until the objective function value is less than the preset threshold. At this point, the current production stage weight coefficient, production process weight coefficient, and production variable weight coefficient are used as the optimal model parameters. Identify which stages, processes, and variables have a critical impact on product quality (e.g., surface smoothness).

[0052] Specifically, the step of calculating the optimal production stage weight coefficients for the current iteration using coordinate descent includes calculating the production process weight coefficients. and production variable weighting coefficients These are treated as known constants and substituted into the objective function. Through this substitution, the originally complex objective function concerning the coupling of three levels of coefficients is transformed into a single-variable convex optimization subproblem concerning only the production stage weight coefficients. This subproblem is then solved to find the value of the overall objective function, including the prediction error term and the sparse penalty term for the stage coefficients, which is taken as the optimal solution for the production stage weight coefficients in the current iteration. Similarly, the process of calculating the optimal production process weight coefficients and production variable weight coefficients in the current iteration using coordinate descent is similar to the process described above.

[0053] Specifically, the method proposed in this application runs in real time on the production line, periodically updating the identification results to provide decision support for on-site engineers (such as adjusting kiln temperature control and optimizing fan settings). Through this method, enterprises have achieved improved product qualification rates on different production lines in actual production.

[0054] Example 3 This embodiment proposes a key production factor identification system for multi-stage product manufacturing. In this embodiment, the system is used to implement a key production factor identification method for multi-stage product manufacturing. A schematic diagram is shown below. Figure 3 As shown, it includes: The production variable acquisition unit is used to collect production variables in real time along the production line of a multi-stage product production line. The production variables include continuous production variables and discrete production variables. The time lag calculation unit is used to calculate the time lag of the product from each production stage to the final inspection position in multi-stage product production, based on the conveyor speed of the production line, using the inspection time of the product at the final inspection position as the global time anchor point. The production factor dataset construction unit is used to integrate continuous and discrete production variables based on time lag, and process the continuous production variables to form a production factor dataset corresponding to the product. The key production factor solution unit is used to construct and train a hierarchical sparse partial function linear regression model. The integrated production factor dataset is input into the trained hierarchical sparse partial function linear regression model and solved to obtain the key production factors.

[0055] Example 4 In this embodiment, a computer device 100 is proposed, which includes a memory 101, a processor 102, and a computer program stored in the memory 101 that can be executed by the processor. The processor 102 executes the computer program to implement a method for identifying key production factors for multi-stage product manufacturing. A schematic diagram of the device is shown below. Figure 4 As shown.

[0056] In this embodiment, a computer storage medium is also proposed, on which a computer program is stored. The computer program includes program instructions, which, when executed by a computer, cause the computer to perform a method for identifying key production factors for multi-stage product manufacturing.

[0057] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying key production factors in multi-stage product manufacturing, characterized in that, The key production factors include: production stage, production process, and production variables; the method includes the following steps: In a multi-stage product production line, production variables for each production stage are collected in real time along the production line. These production variables include continuous production variables and discrete production variables. Using the inspection time of the product at the final inspection position as the global time anchor point, and based on the conveying speed of the production line, the time lag of the product from each production stage to the final inspection position is calculated in multi-stage product production. Based on time lag, continuous production variables and discrete production variables are integrated separately, and the continuous production variables are processed to form a production factor dataset corresponding to the product. A hierarchical sparse partial function linear regression model is constructed and trained. The integrated production factor dataset is then input into the trained hierarchical sparse partial function linear regression model to obtain the key production factors.

2. The method for identifying key production factors for multi-stage product manufacturing according to claim 1, characterized in that, When collecting production variables at each stage of production in real time along the production line, the sensors deployed at each production stage are used to collect the production variables of the product at each production stage. The process of calculating the time lag from each production stage to the final inspection position is as follows: Based on the inspection time of the product at the final inspection location Construct global time anchors; Obtain the physical transmission distance between each production stage and the final inspection location. , Indicates the production stage sequence, based on the physical transmission distance. Conveyor speed of the production line in the corresponding production stage Calculate the time lag of the product at each stage of production. ; Based on the time lag of products at different stages of production and global time anchor Calculate the time lag of the product at each stage of production on the production line; the time lag includes the instantaneous lag of discrete production variables. Lag at both ends of continuous production variables .

3. The method for identifying key production factors for multi-stage product manufacturing according to claim 2, characterized in that, The integration of continuous and discrete production variables based on time lag includes: Based on the instantaneous lag of the discrete production variables Lag at both ends of continuous production variables Obtain products At each production stage, observations are conducted, and fragments of discrete and continuous production variables are obtained for each stage.

4. The method for identifying key production factors for multi-stage product manufacturing according to claim 3, characterized in that, The processing of continuous production variables includes: extracting the continuous production variables with lags at both ends based on principal component analysis. Segments within the interval Principal component scores; The production factor dataset corresponding to the product includes: by product Summary of continuous production variables is lagging at both ends The principal component scores and discrete production variables within the interval segments form a production factor dataset.

5. The method for identifying key production factors for multi-stage product manufacturing according to claim 4, characterized in that, The hierarchical sparse partial function linear regression model includes an objective function and constraints. The expression for the objective function is as follows: The expression for the constraint condition is: in, This represents the weighting coefficient for the production stage. Indicates the weighting coefficient of the production process. This represents the weighting coefficients of the production variable. The regularization coefficient represents the weighting factor of the production stage. The regularization coefficient represents the weighting factor of the production process. The regularization coefficient represents the weighting coefficient of the production variable. Indicates the production stage group. Indicates the group of the production process.

6. The method for identifying key production factors for multi-stage product manufacturing according to claim 5, characterized in that, Before training the hierarchical sparse partial function linear regression model, the method further includes: automatically selecting the regularization coefficients of the production stage weight coefficients using the Bayesian information criterion. Regularization coefficient of production process weighting coefficient Regularization coefficients of production variable weights .

7. The method for identifying key production factors for multi-stage product manufacturing according to claim 6, characterized in that, During the training of the hierarchical sparse partial function linear regression model, the production stage weight coefficients in the model are updated alternately using block coordinate descent. Production process weighting coefficient and production variable weighting coefficients .

8. A key production factor identification system for multi-stage product manufacturing, characterized in that, The system is used to implement the method of any one of claims 1 to 7, comprising: The production variable acquisition unit is used to collect production variables in real time along the production line of a multi-stage product production line. The production variables include continuous production variables and discrete production variables. The time lag calculation unit is used to calculate the time lag of the product from each production stage to the final inspection position in multi-stage product production, based on the conveyor speed of the production line, using the inspection time of the product at the final inspection position as the global time anchor point. The production factor dataset construction unit is used to integrate continuous and discrete production variables based on time lag, and process the continuous production variables to form a production factor dataset corresponding to the product. The key production factor solution unit is used to construct and train a hierarchical sparse partial function linear regression model. The integrated production factor dataset is input into the trained hierarchical sparse partial function linear regression model and solved to obtain the key production factors.

9. A computer device, characterized in that, The computer device includes a memory, a processor, and a computer program stored in the memory that can be run on the processor, wherein the processor executes the computer program to implement the method according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that, It stores a computer program, which includes program instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1 to 7.

Images