CN117391641A

CN117391641A - Pilatory production flow management method and system

Info

Publication number: CN117391641A
Application number: CN202311697952.9A
Authority: CN
Inventors: 黄亮
Original assignee: Zhuhai Xingzhi Biotechnology Co ltd
Current assignee: Zhuhai Xingzhi Biotechnology Co ltd
Priority date: 2023-12-12
Filing date: 2023-12-12
Publication date: 2024-01-12

Abstract

The invention relates to the technical field of production process automation, in particular to a pilatory production process management method and system, comprising the following steps: based on production data, adopting an ETL data fusion technology and a data cleaning and standardization method to integrate and preprocess data so as to generate a standardized data set. In the invention, a convolutional neural network and a recurrent neural network are used for carrying out nonlinear relation analysis on a standardized data set, and complex association between ground observation data is deeply realized, so that the decision process is more accurate, model parameters are adaptively adjusted by a reinforcement learning algorithm and a feedback circulation mechanism, the practical value of the model is enhanced, a decision tree algorithm and a rule engine optimize the production flow, a genetic algorithm and a linear programming optimize resource allocation and flow adjustment, the flexibility and resource efficiency of the production line are enhanced, and the application of a real-time data flow processing technology and an isolated forest algorithm provide real-time monitoring and anomaly detection for the production flow, and the stability and safety of the production process are improved.

Description

Pilatory production flow management method and system

Technical Field

The invention relates to the technical field of production flow automation, in particular to a germicide production flow management method and system.

Background

The field of production process automation technology relates to the use of automation technology to manage and optimize production processes. The core of the technology is to reduce human intervention and improve efficiency and production quality. Automated processes typically include the use of sensors, control systems, software, and machine learning algorithms to monitor and adjust the production process. In particular applications such as the production of hair growth agents, this means that each step from raw material handling to packaging is controlled and optimised by an automated system, ensuring consistency of product quality while reducing production costs and time.

The pilatory production flow management method refers to a set of flow management system specially designed for pilatory manufacture. The main purpose of this method is to ensure the production process of the hair tonic to be efficient and stable, and to maintain the quality of the product. By this management method, the manufacturer can better control the use of raw materials, optimize the production steps, reduce waste, and ensure that the final product meets safety and performance standards. The production efficiency is improved, the consistency of the product quality is maintained, the supervision requirement is met, and the trust of consumers to the product is improved. To achieve these goals, germinant production process management methods typically employ various automated and information technology approaches. This includes the use of advanced production equipment, automated control systems, real-time monitoring techniques, and data analysis tools. With these tools, the production flow can be monitored and adjusted in real time to cope with any factors that may affect the production efficiency and product quality. For example, by real-time data analysis, the production parameters can be adjusted in real-time to ensure that each batch of hair tonic meets the predetermined criteria. In addition, the production flow can be further optimized, potential problems can be predicted, and higher-level automation can be realized by adopting machine learning and artificial intelligence technology.

The traditional germinal agent production flow management method has some defects. First, conventional methods often do not have the ability to handle complex nonlinear relationships in relational analysis, which limits the deep understanding of the production process. In addition, the model adjustment is based on experience judgment, and reinforcement learning and self-adaptive adjustment mechanisms are lacked, so that the optimization effect is limited. In terms of production flow optimization, the conventional method lacks efficient decision support tools, so that the optimization scheme is often not accurate and efficient enough. The resource allocation and the flow adjustment also depend on manual experience, lack of algorithm support, and have low efficiency. Finally, traditional methods often suffer from slow response in terms of production monitoring and anomaly detection, and lack real-time monitoring capabilities, which reduce the stability and safety of the production process. Therefore, the traditional method has larger limitations in the aspects of data processing, relation analysis, model optimization, flow optimization, resource allocation, anomaly detection and the like.

Disclosure of Invention

The invention aims to solve the defects in the prior art, and provides a method and a system for managing the production flow of a germinal agent.

In order to achieve the above purpose, the present invention adopts the following technical scheme: a method for managing the production flow of a germinal agent comprises the following steps:

S1: based on production data, adopting an ETL data fusion technology and a data cleaning and standardization method to integrate and preprocess data so as to generate a standardized data set;

s2: based on the standardized data set, a convolutional neural network and a recurrent neural network are adopted to perform nonlinear relation analysis, and a data relation analysis report is generated;

s3: based on the data relation analysis report, adopting a reinforcement learning algorithm and a feedback circulation mechanism to carry out model parameter self-adaptive adjustment to generate an optimized analysis model;

s4: based on the optimized analysis model, using a decision tree algorithm and a rule engine to generate a production flow optimization suggestion, so as to obtain a production flow optimization scheme;

s5: according to the production flow optimization scheme, adopting a genetic algorithm and linear programming to allocate resources and adjust the flow of the production line, and completing the optimization adjustment of the production line;

s6: based on the production line optimization adjustment, adopting a real-time data flow processing technology and an isolated forest algorithm to perform real-time monitoring and anomaly detection on the production flow, and generating an anomaly detection report;

s7: based on the anomaly detection report and the historical data, adopting a simulated annealing algorithm or Bayesian optimization to perform iterative optimization and feedback adjustment, and establishing a continuous optimization scheme of the production flow;

The standardized data set comprises integrated raw material characteristic data, environment condition data and equipment performance data, the data relation analysis report specifically comprises nonlinear relation and potential pattern analysis among key data, the optimized analysis model specifically comprises an adjusted deep learning model parameter and structure, the production process optimization scheme comprises a raw material proportioning adjustment scheme and a production link operation parameter change scheme, the production line optimization adjustment specifically comprises a resource reallocation scheme and a production sequence adjustment strategy, the anomaly detection report comprises anomaly pattern recognition and cause analysis, and the continuous optimization scheme of the production process specifically comprises long-term trend analysis and improvement point recognition.

As a further scheme of the invention, based on production data, an ETL data fusion technology and a data cleaning and standardization method are adopted to integrate and preprocess data, and the step of generating a standardized data set is specifically as follows:

s101: based on the production data, performing data grabbing and sorting by adopting a data extraction technology to generate a preliminary extraction data set;

s102: based on the preliminary extraction data set, applying a data cleaning and format standardization technology to carry out data deduplication, missing value processing, outlier detection and unified data format, and generating a cleaning and standardization data set;

S103: based on the cleaning and standardized data set, importing the processed data into a data warehouse by using a data loading method to generate an integrated data warehouse;

s104: based on the integrated data warehouse, performing data preprocessing, performing feature scaling and single-heat coding, and generating a standardized data set;

the data extraction technology comprises SQL query and API call, the data cleaning comprises regular expression and data conversion function, the data loading method comprises batch uploading and increment synchronization, and the data preprocessing comprises using normalization technology and transcoding strategy.

As a further aspect of the present invention, the step of performing nonlinear relation analysis and generating a data relation analysis report using a convolutional neural network and a recurrent neural network based on the standardized data set specifically includes:

s201: based on the standardized data set, adopting a convolutional neural network to process the feature extraction of the image and the time sequence data, and generating a CNN processing result;

s202: based on the CNN processing result, analyzing the time dependency relationship in the sequence data by using a recurrent neural network to generate an RNN analysis result;

s203: performing data fusion and relation mapping, and combining the CNN processing result and the RNN analysis result to generate a comprehensive relation analysis result;

S204: based on the comprehensive relationship analysis result, compiling a data relationship analysis report, including key discovery and pattern recognition, and generating a data relationship analysis report;

the convolutional neural network comprises a convolutional layer, a pooling layer and a full-connection layer, the recurrent neural network comprises a long-term memory network and a gating circulation unit, and the data fusion comprises feature fusion and model integration strategies.

As a further scheme of the invention, based on the data relation analysis report, a reinforcement learning algorithm and a feedback circulation mechanism are adopted to carry out self-adaptive adjustment on model parameters, and the steps for generating an optimized analysis model are as follows:

s301: based on the data relation analysis report, performing model parameter adjustment by adopting a reinforcement learning algorithm to generate a preliminary adjustment model;

s302: based on the preliminary adjustment model, performing simulated environment feedback training to generate an interactive training model;

s303: based on the interactive training model, an adaptive learning rate adjustment strategy is applied to generate a parameter optimization model;

s304: based on the parameter optimization model, adopting a continuous performance feedback loop to generate an optimized analysis model;

the reinforcement learning algorithm comprises a deep Q network and a strategy gradient method, the simulation environment feedback training specifically comprises the steps of establishing a virtual environment to carry out strategy iteration, the self-adaptive learning rate adjustment strategy comprises the step of using an Adam optimizer and a learning rate attenuation technology, and the continuous performance feedback loop comprises periodic evaluation and real-time adjustment.

As a further scheme of the invention, based on the optimized analysis model, a decision tree algorithm and a rule engine are used for generating a production flow optimization proposal, and the steps for obtaining the production flow optimization scheme are as follows:

s401: based on the optimized analysis model, performing preliminary flow analysis by using a decision tree algorithm to generate a preliminary decision tree analysis result;

s402: based on the analysis result of the preliminary decision tree, rule extraction and refinement are implemented, operation rules are extracted, and an extraction rule set is generated;

s403: based on the refined rule set, applying rule engine simulation and analysis to optimize a decision process and generating a rule engine analysis result;

s404: based on the rule engine analysis result, formulating a comprehensive production flow optimization proposal to generate a production flow optimization proposal;

the decision tree algorithm comprises a random forest and a gradient lifting decision tree, the rule extraction and refinement involves analyzing key paths and nodes of the decision tree, the rule engine simulation and analysis comprise simulation execution and effect prediction of the rule, and the comprehensive production flow optimization suggestion comprises efficiency lifting, cost reduction and risk management.

As a further scheme of the invention, according to the production flow optimization scheme, a genetic algorithm and linear programming are adopted to carry out resource allocation and flow adjustment on the production line, and the steps for completing the production line optimization adjustment are specifically as follows:

S501: based on the production flow optimization scheme, a genetic algorithm is adopted to conduct preliminary planning on resource allocation, and a resource allocation scheme draft is generated;

s502: based on the resource allocation scheme draft, applying linear programming to carry out scheme adjustment to generate a resource allocation scheme;

s503: based on the resource allocation scheme, carrying out flow adjustment simulation of the production line to generate a flow adjustment simulation result;

s504: based on the flow adjustment simulation result, carrying out actual optimization adjustment of the production line to generate optimization adjustment of the production line;

the genetic algorithm comprises a coding strategy, a fitness function design, a cross operation and a mutation operation, the linear programming comprises an objective function construction and constraint condition setting, the production line flow adjustment simulation uses a discrete event simulation technology to verify the performance under the differential configuration, and the actual optimization adjustment of the production line is specifically the actual adjustment of physical layout and operation flow according to simulation feedback.

As a further scheme of the invention, based on the optimized adjustment of the production line, a real-time data flow processing technology and an isolated forest algorithm are adopted to carry out real-time monitoring and anomaly detection on the production flow, and the steps for generating an anomaly detection report are specifically as follows:

S601: based on the optimized adjustment result of the production line, a real-time data stream processing technology is deployed to monitor the production process, and a real-time monitoring system is generated;

s602: based on the real-time monitoring system, data preprocessing and feature extraction are carried out to generate preprocessed data;

s603: based on the preprocessing data, performing abnormal mode identification by adopting an isolated forest algorithm, and generating an abnormal detection preliminary result;

s604: based on the initial result of abnormality detection, carrying out deep analysis and cause exploration of an abnormality mode to generate an abnormality detection report;

the real-time data stream processing technique includes Apache Kafka and Apache Spark Streaming, the preprocessing includes data cleaning, normalization and feature extraction based on time windows, and the isolated forest algorithm is used for anomaly detection of a multi-dimensional dataset.

As a further scheme of the invention, based on the anomaly detection report and the historical data, adopting a simulated annealing algorithm or Bayesian optimization to perform iterative optimization and feedback adjustment, and establishing a continuous optimization scheme of the production flow specifically comprises the following steps:

s701: based on the anomaly detection report and the historical data, performing solution space search by adopting a simulated annealing algorithm, gradually reducing characteristics to find a global optimal solution, and generating a preliminary optimization scheme;

S702: based on the preliminary optimization scheme, generating a further optimization scheme by adopting a parameter combination with optimal Bayesian optimization prediction performance;

s703: based on the advanced optimization scheme, a feedback adjustment mechanism is applied, real-time adjustment of optimization parameters is carried out according to production data, and a dynamic adjustment scheme is generated;

s704: based on the dynamic adjustment scheme, comprehensive evaluation and fine adjustment are executed, the effectiveness of the scheme is verified and ensured, and a continuous optimization scheme of the production flow is generated;

the simulated annealing algorithm comprises initial temperature setting, probability solution selection and annealing process control, the Bayesian optimization comprises Gaussian process modeling, priori information utilization and posterior probability optimization, the feedback adjustment mechanism comprises real-time data analysis, effect evaluation and parameter adjustment, and the comprehensive evaluation and fine adjustment comprises scheme performance evaluation, risk analysis and parameter fine adjustment.

The system comprises a data integration module, a relation analysis module, a model optimization module, a flow optimization suggestion module, a resource allocation module and a real-time monitoring and abnormality detection module.

As a further scheme of the invention, the data integration module performs data integration and preprocessing by adopting an ETL technology and combining a data cleaning and standardization method based on production data to generate a standardized data set;

the relation analysis module is used for processing the image and the time sequence data by adopting a convolutional neural network based on a standardized data set and generating a data relation analysis report by combining a recurrent neural network to analyze the time dependency relation;

the model optimization module is used for carrying out model parameter self-adaptive adjustment by adopting a reinforcement learning algorithm based on the data relation analysis report to generate an optimized analysis model;

the flow optimization suggestion module optimizes a decision process by using a decision tree algorithm and a rule engine based on the optimized analysis model to generate a production flow optimization scheme;

the resource allocation module performs resource allocation and flow adjustment on the production line by adopting a genetic algorithm and linear programming according to a production flow optimization scheme to complete the optimization adjustment of the production line;

the real-time monitoring and anomaly detection module is used for carrying out real-time monitoring and anomaly detection on the production flow by adopting a real-time data flow processing technology and an isolated forest algorithm based on the production line optimization adjustment result, and generating an anomaly detection report.

Compared with the prior art, the invention has the advantages and positive effects that:

according to the method, the convolutional neural network and the recurrent neural network are used for carrying out nonlinear relation analysis on the standardized data set, so that complex association between the ground observation data can be further achieved, and the decision process is more accurate. And the self-adaptive adjustment of the model parameters is carried out by applying a reinforcement learning algorithm and a feedback circulation mechanism, so that the analysis model is more close to the actual production requirement, and the application value of the model is improved. The application of the decision tree algorithm and the rule engine then helps to generate a more rational and efficient production flow optimization scheme. And the production line is subjected to resource allocation and flow adjustment through a genetic algorithm and linear programming, so that the flexibility and the resource allocation efficiency of the production line are enhanced. The application of the real-time data flow processing technology and the isolated forest algorithm provides real-time monitoring and anomaly detection for the production process, and improves the stability and safety of the production process.

Drawings

FIG. 1 is a schematic workflow diagram of the present invention;

FIG. 2 is a S1 refinement flowchart of the present invention;

FIG. 3 is a S2 refinement flowchart of the present invention;

FIG. 4 is a S3 refinement flowchart of the present invention;

FIG. 5 is a S4 refinement flowchart of the present invention;

FIG. 6 is a S5 refinement flowchart of the present invention;

FIG. 7 is a S6 refinement flowchart of the present invention;

FIG. 8 is a S7 refinement flowchart of the present invention;

fig. 9 is a system flow diagram of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

In the description of the present invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate orientations or positional relationships based on the orientation or positional relationships shown in the drawings, merely to facilitate describing the present invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, in the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.

Embodiment one: referring to fig. 1, the present invention provides a technical solution: a method for managing the production flow of a germinal agent comprises the following steps:

s2: based on the standardized data set, adopting a convolutional neural network and a recurrent neural network to perform nonlinear relation analysis and generating a data relation analysis report;

s3: based on the data relation analysis report, adopting a reinforcement learning algorithm and a feedback circulation mechanism to carry out self-adaptive adjustment on model parameters so as to generate an optimized analysis model;

s6: based on the optimized adjustment of the production line, adopting a real-time data flow processing technology and an isolated forest algorithm to monitor the production flow in real time and detect the abnormality to generate an abnormality detection report;

the standardized data set comprises integrated raw material characteristic data, environment condition data and equipment performance data, the data relation analysis report specifically comprises nonlinear relation and potential mode analysis among key data, the optimized analysis model specifically comprises an adjusted deep learning model parameter and structure, the production flow optimization scheme comprises a raw material proportion adjustment scheme and a production link operation parameter change scheme, the production line optimization adjustment specifically comprises a resource redistribution scheme and a production sequence adjustment strategy, the anomaly detection report comprises anomaly mode identification and cause analysis, and the continuous optimization scheme of the production flow specifically comprises long-term trend analysis and improvement point identification.

Through ETL data fusion technology and data cleaning standardization, data quality and consistency are ensured, accuracy and decision speed of data analysis are improved, and overall production efficiency is further improved. And secondly, the convolutional neural network and the recurrent neural network are used for carrying out deep analysis, so that the accuracy of the production process is improved, errors are reduced, and the product quality is improved. And thirdly, the model parameters are adaptively adjusted through reinforcement learning and feedback circulation mechanisms, so that the flexibility and adaptability of the production process are improved, and the production efficiency and the stability of the product quality are maintained. In addition, the application of the decision tree algorithm and the rule engine optimizes the production flow, efficiently utilizes resources and reduces the cost. Genetic algorithms and linear programming make production line resource allocation and process adjustment more scientific and efficient. Real-time data flow processing and isolated forest algorithm realize accurate and timely production monitoring and anomaly detection, reduce production interruption risk and guarantee flow continuity. Finally, the simulated annealing algorithm or Bayesian optimization ensures continuous improvement of the production flow, and long-term production efficiency and market response capability are improved.

Referring to fig. 2, based on production data, the steps of integrating and preprocessing data by adopting the ETL data fusion technology and the data cleaning and standardization method to generate a standardized data set are specifically as follows:

s103: based on the cleaning and standardized data sets, importing the processed data into a data warehouse by using a data loading method to generate an integrated data warehouse;

s104: based on the integrated data warehouse, carrying out data preprocessing, carrying out characteristic scaling and single-heat coding, and generating a standardized data set;

the data extraction technology comprises SQL query and API call, the data cleaning comprises regular expression and data conversion function, the data loading method comprises batch uploading and increment synchronization, and the data preprocessing comprises normalization technology and transcoding strategy.

In S101, the source of the data is identified and the appropriate database connection or API key is configured. Related data is extracted and consolidated into a preliminary extracted data set using SQL query or API call techniques. Through SQL query, the required data table and field are selected, and the result is saved in a temporary table. At the same time, through API calls, the data is retrieved and saved in an appropriate data structure, such as a JSON or CSV format.

In S102, the data is subjected to deduplication processing, and any duplicate records are deleted. The missing values are then processed and a packed mean, median, or other interpolation method may be employed. Abnormal value detection is carried out, a statistical method or a machine learning model is used for identifying the abnormality, and corresponding processing is carried out, wherein the abnormality can be deleted, replaced or marked as the abnormality. Finally, through regular expressions and data conversion functions, the data is ensured to conform to the consistent format and standard.

In S103, the data warehouse structure is designed, and the relationships and field definitions of the tables are specified. The processed data is then imported into the data warehouse in batches by batch upload. For data needing to be updated in real time, an incremental synchronization function is introduced to ensure that the data in the data warehouse is kept up to date.

In S104, further data preprocessing is performed in the integrated data warehouse to meet modeling and analysis requirements. Feature scaling is performed to ensure consistent numerical ranges for the different features, and common methods include Min-Max scaling or Z-score normalization. The classification variables are unithermally encoded and converted to binary vectors for use in a machine learning model.

Referring to fig. 3, based on the standardized data set, the steps of performing nonlinear relation analysis by adopting a convolutional neural network and a recurrent neural network and generating a data relation analysis report specifically include:

s203: performing data fusion and relation mapping, and combining a CNN processing result and an RNN analysis result to generate a comprehensive relation analysis result;

s204: based on the comprehensive relation analysis result, compiling a data relation analysis report, including key discovery and pattern recognition, and generating a data relation analysis report;

In S201, in a stage of processing the normalized data set, a Convolutional Neural Network (CNN) is first used to perform feature extraction, which is applicable to image and time series data. The CNN model is configured and trained, including a convolutional layer, a pooling layer, and a fully-connected layer. The convolution layer is used for capturing local features in the data set, the pooling layer reduces dimensionality, and the full connection layer integrates the features. And (3) applying the model to the standardized data set through training to generate CNN processing results. These results reflect the spatial and temporal characteristics in the dataset, laying the foundation for further analysis.

In S202, based on the CNN processing result, a Recurrent Neural Network (RNN) is introduced to analyze the time-dependent relationship in the sequence data. The RNN model is configured to include a long-short-term memory network (LSTM) and a gate loop unit (GRU) to capture a time-series pattern of data. Training is carried out, CNN processing results are used as input sequences, and RNN analysis results are obtained. This step helps to reveal the evolution law of the data over time, further deepening the understanding of the nonlinear relationship.

In S203, the CNN processing result and the RNN analysis result are subjected to data fusion, which is implemented through feature fusion and model integration strategies. Feature fusion may be a simple join operation or a weighted average, while model integration may employ voting, averaging, or stacking, among others. The method aims at synthesizing information from different neural networks and generating a more comprehensive relation analysis result. This step emphasizes the advantage of integrating multiple models, improving the understanding of complex data relationships.

In S204, a data relationship analysis report is created based on the comprehensive relationship analysis result. The report covers key findings and pattern recognition including non-linear relationships, trends, periodicity, etc. The structure of the report should be clear, including descriptive statistics, visual charts, and model output interpretation. And simultaneously, the key results are interpreted and verified to ensure the accuracy and the credibility of analysis. This step is a key step in converting the output of the deep learning model into actual business insight.

Referring to fig. 4, based on a data relationship analysis report, a reinforcement learning algorithm and a feedback loop mechanism are adopted to perform model parameter self-adaptive adjustment, and the steps of generating an optimized analysis model are specifically as follows:

In S301, model parameter adjustment is performed using a reinforcement learning algorithm based on the previous data relationship analysis report. The deep Q network and the policy gradient method are chosen, which enable complex policies to be learned in different contexts. The goal of reinforcement learning is to maximize the performance of the model by optimizing its weights and superparameters. Through the step, a preliminary adjustment model is generated, and a foundation is provided for model optimization in the next step.

In S302, based on the preliminary adjustment model, a feedback training of the simulation environment is performed. The method comprises the steps of establishing a virtual environment, simulating an actual application scene, and carrying out model training through strategy iteration. In a virtual environment, the model interacts with the environment, and the strategy is continuously adjusted through feedback so as to adapt to different data distribution and change. The generated interactive training model has better robustness and generalization capability, and can be better adapted to actual application scenes.

In S303, an adaptive learning rate adjustment strategy is introduced by means of an interactive training model. This strategy uses Adam optimizers in combination with learning rate decay techniques. Adam optimizers are adaptive, while learning rate decay techniques help flexibly adjust learning rates during different phases of training to better accommodate changes in data distribution. Through the step, the convergence speed and performance of the model are further improved.

In S304, based on the parameter optimization model, a continuous performance feedback loop is employed to ensure that the model maintains a high level of performance in practical applications. This feedback loop includes periodic evaluation and real-time adjustment. And (3) performing model performance evaluation regularly, and comparing with service requirements to ensure that the model is still effective. Simultaneously, the model parameters are adjusted in real time to adapt to the change of the real-time data. This cycle ensures that the model remains highly adaptable in a constantly changing environment, improving the robustness and sustainability of the model.

Referring to fig. 5, based on the optimized analysis model, using a decision tree algorithm and a rule engine, the production flow optimization suggestion is generated, and the steps for obtaining the production flow optimization scheme specifically include:

s401: based on the optimized analysis model, performing primary flow analysis by using a decision tree algorithm to generate a primary decision tree analysis result;

s403: based on the refined rule set, applying rule engine simulation and analysis to optimize the decision process and generate rule engine analysis results;

s404: based on the analysis result of the rule engine, formulating a comprehensive production flow optimization proposal to generate a production flow optimization proposal;

the decision tree algorithm comprises a random forest and gradient lifting decision tree, the rule extraction and refinement relate to analyzing key paths and nodes of the decision tree, the rule engine simulation and analysis comprise simulation execution and effect prediction of rules, and the comprehensive production flow optimization suggestion comprises efficiency lifting, cost reduction and risk management.

In S401, the decision tree algorithm is applied to the preliminary flow analysis of the optimized analysis model by using the decision tree algorithm including random forest and gradient lifting decision tree. Through learning the data, the algorithm generates a preliminary decision tree analysis result, and key decision nodes and correlations thereof in the production flow are revealed. The goal of this step is to build a base framework that provides the basis for subsequent rule extraction.

In S402, rule extraction and refinement are performed based on the preliminary decision tree analysis result. This includes analyzing critical paths and nodes of the decision tree, extracting operational rules, and generating a refined rule set. This rule set will contain critical rules in the production flow, providing strong support for the application of the rule engine. The goal of this step is to mine the actual operational rules from the structure of the decision tree.

In S403, the rule engine simulates and analyzes the refined rule set by applying the rule engine at this step. This involves simulated execution of rules and effect prediction. Interactions between rules and their impact on the production flow are better understood through simulation of the rules engine. The goal of this step is to optimize the decision process, ensuring that the actual execution of the rules can bring about the desired effect.

In S404, the integrated production process optimization suggestion is generated to formulate an integrated production process optimization suggestion based on the rule engine analysis result. These suggestions will include comprehensive optimization schemes in terms of efficiency improvement, cost reduction, and risk management. By comprehensively considering the analysis results of the rule engine, the generated production flow optimization scheme is ensured to be comprehensive and feasible, and can be effectively implemented in actual operation.

Referring to fig. 6, according to the production flow optimization scheme, a genetic algorithm and a linear programming are adopted to allocate resources and adjust the flow of the production line, and the steps for completing the optimization and adjustment of the production line are specifically as follows:

s501: based on a production flow optimization scheme, a genetic algorithm is adopted to conduct preliminary planning on resource allocation, and a resource allocation scheme draft is generated;

s502: based on the draft of the resource allocation scheme, applying linear programming to carry out scheme adjustment to generate the resource allocation scheme;

s503: based on the resource allocation scheme, carrying out flow adjustment simulation on the production line to generate a flow adjustment simulation result;

s504: based on the flow adjustment simulation result, carrying out actual optimization adjustment of the production line to generate the optimization adjustment of the production line;

the genetic algorithm comprises a coding strategy, fitness function design, cross operation and mutation operation, the linear programming comprises objective function construction and constraint condition setting, the production line flow adjustment simulation uses a discrete event simulation technology to verify the performance under the differential configuration, and the actual optimization adjustment of the production line is specifically the actual adjustment of physical layout and operation flow according to simulation feedback.

In S501, on the basis of the production process optimization scheme, a genetic algorithm is adopted to perform preliminary planning on the resources. This includes designing coding strategies, constructing fitness functions, and defining crossover and mutation operations. A preliminary draft of resource allocation schemes is generated through genetic algorithm, aiming at exploring potential optimization schemes.

In S502, based on the resource allocation scheme draft, a linear program is introduced to perform scheme adjustment. This step involves the construction of the objective function and the setting of constraints to more finely adjust the allocation of resources. The application of linear programming aims at optimizing the resource allocation and ensuring that the production line runs in an optimal state. The generated resource allocation scheme is adjusted by the step, so that the method is more in line with the actual situation and the maximum optimization target.

In S503, based on the final resource allocation scheme, a discrete event simulation technique is used to perform a line flow adjustment simulation. This step involves verifying performance under differential configuration, and evaluating the efficiency and stability of the production line under new configuration by simulating different production scenarios. The results of the simulation will provide key performance indicators, providing guidance for the actual adjustment.

In S504, based on the result of the flow adjustment simulation, actual optimization adjustment of the production line is performed. This step includes making the actual adjustments to the physical layout and operational flow based on the analog feedback. By applying the simulated best practices to the actual operation, the production line is enabled to achieve the best performance under the new optimal configuration. This involves the adjustment of the actual operational level of equipment placement, staff scheduling, etc.

Referring to fig. 7, based on the optimization adjustment of the production line, the real-time monitoring and anomaly detection of the production process are performed by adopting a real-time data flow processing technology and an isolated forest algorithm, and the steps of generating an anomaly detection report are specifically as follows:

s602: based on a real-time monitoring system, data preprocessing and feature extraction are carried out to generate preprocessed data;

s604: based on the initial result of the anomaly detection, carrying out the deep analysis and the cause exploration of the anomaly mode to generate an anomaly detection report;

real-time data stream processing techniques include Apache Kafka and Apache Spark Streaming, preprocessing including data cleaning, normalization and time window based feature extraction, and isolated forest algorithms for anomaly detection of multi-dimensional datasets.

In S601, after the real-time data stream processing technology is deployed to optimize the production process monitoring line, the Apache Kafka is used to collect the data on the production line. Kafka is a distributed stream processing system for processing real-time data sources. Next, the real-time data collected by Kafka was processed using Apache Spark Streaming. The following are code samples of Kafka and Spark Streaming integration:

from pyspark.streaming import StreamingContextfrom pyspark.streaming.kafka import KafkaUtils

sparkConf = SparkConf().setAppName('KafkaStreamProcessing')

ssc=streamingcontext (sparkConf, 5) # 5 seconds batch

kafkaStream = KafkaUtils.createStream(ssc, 'localhost:2181', 'raw-event-streaming-consumer', {'raw-event':1})

In S602, data preprocessing and feature extraction are performed. Once there is a real-time data stream, the data needs to be pre-processed, including cleaning, normalization, and feature extraction. The data cleaning comprises the steps of eliminating null values and abnormal values, normalizing the values to a certain interval range, and extracting features to extract useful data or data subsets helping to provide better information from the original data. The following is a simple example of data preprocessing:

from sklearn.preprocessing import MinMaxScaler

def preprocess_data(stream):

data cleansing #

stream = stream.na.drop()

Feature extraction, e.g. extracting average, maximum and minimum values

features = stream.window(60).aggregate(

{

'value':'avg',

'value': 'min',

'value': 'max'

}

)

Data normalization

scaler = MinMaxScaler()

features = scaler.fit_transform(features)

return features

In S603, an isolated forest algorithm is used to perform abnormality pattern recognition, and the data generated in the preprocessing step is used to perform abnormality detection. The isolated forest algorithm is a very efficient anomaly detection method. In real-time data stream processing, abnormal data is effectively identified. The following is a simple use example of an orphan forest algorithm:

from sklearn.ensemble import IsolationForest

iso_forest = IsolationForest(n_estimators=100)

iso_forest.fit(features)

# detection abnormality

pred = iso_forest.predict(features)

In S604, the depth analysis and the cause exploration of the abnormal pattern are performed, and the cause of the abnormality is found out as much as possible by performing the depth analysis along with the result of the abnormal pattern recognition. The method comprises the steps of statistical analysis, cause tracing, visual analysis and the like. And in particular will vary depending on the business situation and requirements.

Referring to fig. 8, based on the anomaly detection report and the historical data, the iterative optimization and the feedback adjustment are performed by adopting a simulated annealing algorithm or bayesian optimization, and the steps of establishing a continuous optimization scheme of the production flow are specifically as follows:

s701: based on the anomaly detection report and the historical data, performing solution space search by adopting a simulated annealing algorithm, gradually reducing the characteristics to find a global optimal solution, and generating a preliminary optimization scheme;

s704: based on the dynamic adjustment scheme, comprehensive evaluation and fine adjustment are executed, the validity of the scheme is verified and ensured, and a continuous optimization scheme of the production flow is generated;

the simulated annealing algorithm comprises initial temperature setting, probability solution selection and annealing process control, bayesian optimization comprises Gaussian process modeling, priori information utilization and posterior probability optimization, the feedback adjustment mechanism comprises real-time data analysis, effect evaluation and parameter adjustment, and the comprehensive evaluation and fine adjustment comprises scheme performance evaluation, risk analysis and parameter fine adjustment.

In S701, a solution space search is performed using a simulated annealing algorithm based on the anomaly detection report and the historical data. Firstly, setting an initial temperature, selecting a solution under the guidance of probability, and gradually reducing the characteristics through an annealing process to find a globally optimal solution. This step creates a preliminary optimization scheme that traverses the knowledge space as much as possible through the search process in order to find the optimal combination of parameters.

In S702, based on the preliminary optimization scheme, bayesian optimization is adopted to predict the parameter combination with optimal performance. This step includes gaussian process modeling, a priori information utilization, and posterior probability optimization. Through Bayesian optimization, a advanced optimization scheme is generated, and parameters are adjusted more finely, so that the performance is improved, and the search range is reduced.

In S703, a feedback adjustment mechanism is introduced based on the advanced optimization scheme, and real-time adjustment of the optimization parameters is performed by using the real-time production data. This step includes real-time data analysis, effect assessment and parameter adjustment. Parameters of the optimization scheme are adjusted by continuously analyzing actual production data so as to adapt to changes of production environments, and a dynamic adjustment scheme is generated.

In S704, based on the dynamic adjustment scheme, comprehensive evaluation and fine adjustment are performed, verifying and ensuring the validity of the scheme. This includes solution performance assessment, risk analysis, and parameter tuning. And confirming the effect of the optimization scheme in the actual environment through comprehensive evaluation, and analyzing risks. The fine tuning operation further optimizes the parameters, ensuring continued adaptability and performance of the solution.

Referring to fig. 9, a germinal agent production process management system is used for executing the germinal agent production process management method, and the system includes a data integration module, a relationship analysis module, a model optimization module, a process optimization suggestion module, a resource allocation module, and a real-time monitoring and anomaly detection module.

The data integration module is used for integrating and preprocessing data by adopting an ETL technology and combining a data cleaning and standardization method based on production data to generate a standardized data set;

the relation analysis module is used for processing the image and the time sequence data by adopting a convolutional neural network based on the standardized data set, and generating a data relation analysis report by combining the recurrent neural network to analyze the time dependency relation;

the model optimization module performs model parameter self-adaptive adjustment by adopting a reinforcement learning algorithm based on the data relation analysis report to generate an optimized analysis model;

the resource allocation module performs resource allocation and flow adjustment on the production line by adopting a genetic algorithm and linear programming according to the production flow optimization scheme, so as to complete the optimization adjustment of the production line;

Through ETL technology and data cleaning standardization, high quality and consistency of data are ensured, and reliable data support is provided for the whole production flow. Secondly, the depth relation analysis capability of the system reveals the complex relation and time dependence in the production process, and enhances the understanding of the production flow, thereby improving the product quality and reducing the production problems. Furthermore, the self-adaptive adjustment function of the model optimization module enables the system to flexibly cope with changes of market and production conditions, and production efficiency and flexibility are maintained. In addition, the intelligent flow decision and resource allocation not only optimize the production efficiency, but also reduce the cost and the resource waste. Finally, the functions of real-time monitoring and anomaly detection obviously reduce the risk of production interruption, and ensure the continuity and stability of the process.

The present invention is not limited to the above embodiments, and any equivalent embodiments which can be changed or modified by the technical disclosure described above can be applied to other fields, but any simple modification, equivalent changes and modification made to the above embodiments according to the technical matter of the present invention will still fall within the scope of the technical disclosure.

Claims

1. A method for managing the production flow of a germinal agent, which is characterized by comprising the following steps:

based on production data, adopting an ETL data fusion technology and a data cleaning and standardization method to integrate and preprocess data so as to generate a standardized data set;

based on the standardized data set, a convolutional neural network and a recurrent neural network are adopted to perform nonlinear relation analysis, and a data relation analysis report is generated;

based on the data relation analysis report, adopting a reinforcement learning algorithm and a feedback circulation mechanism to carry out model parameter self-adaptive adjustment to generate an optimized analysis model;

based on the optimized analysis model, using a decision tree algorithm and a rule engine to generate a production flow optimization suggestion, so as to obtain a production flow optimization scheme;

according to the production flow optimization scheme, adopting a genetic algorithm and linear programming to allocate resources and adjust the flow of the production line, and completing the optimization adjustment of the production line;

based on the production line optimization adjustment, adopting a real-time data flow processing technology and an isolated forest algorithm to perform real-time monitoring and anomaly detection on the production flow, and generating an anomaly detection report;

based on the anomaly detection report and the historical data, adopting a simulated annealing algorithm or Bayesian optimization to perform iterative optimization and feedback adjustment, and establishing a continuous optimization scheme of the production flow;

2. The method according to claim 1, wherein the step of integrating and preprocessing data based on production data by adopting ETL data fusion technology and data cleaning and standardization method to generate standardized data set comprises:

based on the production data, performing data grabbing and sorting by adopting a data extraction technology to generate a preliminary extraction data set;

based on the preliminary extraction data set, applying a data cleaning and format standardization technology to carry out data deduplication, missing value processing, outlier detection and unified data format, and generating a cleaning and standardization data set;

Based on the cleaning and standardized data set, importing the processed data into a data warehouse by using a data loading method to generate an integrated data warehouse;

based on the integrated data warehouse, performing data preprocessing, performing feature scaling and single-heat coding, and generating a standardized data set;

3. The method according to claim 1, wherein the steps of performing nonlinear relation analysis using a convolutional neural network and a recurrent neural network based on the standardized data set, and generating a data relation analysis report are specifically as follows:

based on the standardized data set, adopting a convolutional neural network to process the feature extraction of the image and the time sequence data, and generating a CNN processing result;

based on the CNN processing result, analyzing the time dependency relationship in the sequence data by using a recurrent neural network to generate an RNN analysis result;

performing data fusion and relation mapping, and combining the CNN processing result and the RNN analysis result to generate a comprehensive relation analysis result;

Based on the comprehensive relationship analysis result, compiling a data relationship analysis report, including key discovery and pattern recognition, and generating a data relationship analysis report;

4. The method according to claim 1, wherein the step of generating an optimized analysis model by adaptively adjusting model parameters based on the data relationship analysis report using a reinforcement learning algorithm and a feedback loop mechanism comprises:

based on the data relation analysis report, performing model parameter adjustment by adopting a reinforcement learning algorithm to generate a preliminary adjustment model;

based on the preliminary adjustment model, performing simulated environment feedback training to generate an interactive training model;

based on the interactive training model, an adaptive learning rate adjustment strategy is applied to generate a parameter optimization model;

based on the parameter optimization model, adopting a continuous performance feedback loop to generate an optimized analysis model;

5. The method for managing a germinal agent production flow according to claim 1, wherein the steps of generating a production flow optimization proposal by using a decision tree algorithm and a rule engine based on the optimized analysis model to obtain a production flow optimization scheme are specifically as follows:

based on the optimized analysis model, performing preliminary flow analysis by using a decision tree algorithm to generate a preliminary decision tree analysis result;

based on the analysis result of the preliminary decision tree, rule extraction and refinement are implemented, operation rules are extracted, and an extraction rule set is generated;

based on the refined rule set, applying rule engine simulation and analysis to optimize a decision process and generating a rule engine analysis result;

based on the rule engine analysis result, formulating a comprehensive production flow optimization proposal to generate a production flow optimization proposal;

6. The method for managing the production flow of the germinal agent according to claim 1, wherein the steps of performing resource allocation and flow adjustment on the production line by adopting a genetic algorithm and linear programming according to the production flow optimization scheme, and performing the optimization adjustment on the production line are specifically as follows:

based on the production flow optimization scheme, a genetic algorithm is adopted to conduct preliminary planning on resource allocation, and a resource allocation scheme draft is generated;

based on the resource allocation scheme draft, applying linear programming to carry out scheme adjustment to generate a resource allocation scheme;

based on the resource allocation scheme, carrying out flow adjustment simulation of the production line to generate a flow adjustment simulation result;

based on the flow adjustment simulation result, carrying out actual optimization adjustment of the production line to generate optimization adjustment of the production line;

7. The method for managing a germinal agent production process according to claim 1, wherein the steps of performing real-time monitoring and anomaly detection of the production process by using a real-time data flow processing technique and an isolated forest algorithm based on the optimization adjustment of the production line, and generating an anomaly detection report specifically include:

based on the optimized adjustment result of the production line, a real-time data stream processing technology is deployed to monitor the production process, and a real-time monitoring system is generated;

based on the real-time monitoring system, data preprocessing and feature extraction are carried out to generate preprocessed data;

based on the preprocessing data, performing abnormal mode identification by adopting an isolated forest algorithm, and generating an abnormal detection preliminary result;

based on the initial result of abnormality detection, carrying out deep analysis and cause exploration of an abnormality mode to generate an abnormality detection report;

8. The method for managing a germinal agent production flow according to claim 1, wherein based on the anomaly detection report and the historical data, iterative optimization and feedback adjustment are performed by adopting a simulated annealing algorithm or bayesian optimization, and the step of establishing a continuous optimization scheme of the production flow is specifically:

Based on the anomaly detection report and the historical data, performing solution space search by adopting a simulated annealing algorithm, gradually reducing characteristics to find a global optimal solution, and generating a preliminary optimization scheme;

based on the preliminary optimization scheme, generating a further optimization scheme by adopting a parameter combination with optimal Bayesian optimization prediction performance;

based on the advanced optimization scheme, a feedback adjustment mechanism is applied, real-time adjustment of optimization parameters is carried out according to production data, and a dynamic adjustment scheme is generated;

based on the dynamic adjustment scheme, comprehensive evaluation and fine adjustment are executed, the effectiveness of the scheme is verified and ensured, and a continuous optimization scheme of the production flow is generated;

9. A germinal agent production flow management system, characterized in that the germinal agent production flow management method according to any of claims 1-8, the system comprises a data integration module, a relationship analysis module, a model optimization module, a flow optimization suggestion module, a resource allocation module, a real-time monitoring and anomaly detection module.

10. The germinal agent production flow management system of claim 9, wherein the data integration module performs data integration and preprocessing based on production data by adopting ETL technology and combining data cleaning and normalization methods to generate a standardized data set;