Fly ash production line intelligent monitoring system and method based on multi-source data

Abstract

The invention discloses a fly ash production line intelligent monitoring system and method based on multi-source data, and relates to the technical field of data analys.The fly ash production line intelligent monitoring system comprises the steps that historical monitoring records are recorded, fly ash initial component scores are calculated, the initial grade of fly ash is determined, and feature records are determined; calculating an environment score, determining a characteristic process parameter set of the processing procedure, and establishing a classification model of the processing procedure; a finished product score prediction model and an environment score prediction model are constructed respectively, classification is performed according to label values of feature records, and a minimization objective function is established; based on fly ash real-time components and a preparation target, in combination with a classification model, a label value is judged, corresponding target functions are constructed respectively, target function value evaluation and sorting are performed on candidate processing procedures, and the procedure with the minimum target function value is selected as the optimal processing procedure, so that the fly ash processing safety, the resource utilization rate and the operation and maintenance efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and specifically to an intelligent monitoring system and method for a fly ash production line based on multi-source data. Background Art

[0002] With the in-depth promotion of industrial informatization and intelligent manufacturing, the concept of data-driven production management has been widely applied in many key fields such as environmental governance and hazardous waste treatment. Especially in the high-risk and high-complexity process of fly ash treatment, due to the involvement of various physical and chemical reactions, multi-stage collaborative control, and high attention to environmental compliance requirements; However, the sources of fly ash are complex and the composition varies greatly. Even the fly ash produced by the same incineration plant daily has significant differences in aspects such as moisture content, pH value, and metal content. Different treatment methods have different requirements for the types, ratios, and reaction sequences of additives, and the treatment effect highly depends on the accurate identification of raw material properties and the precise execution of process control; At present, most fly ash treatment production lines still adopt a management method combining segmented control and manual intervention, and cannot dynamically adjust the feeding strategy according to the characteristics of raw materials. In order to meet the actual needs of large fluctuations in raw materials, complex processes, and high control requirements during the fly ash treatment process, an intelligent monitoring system and method for a fly ash production line based on multi-source data are proposed to realize a production line-level monitoring system integrating multi-source perception, intelligent judgment, and automatic control, thereby improving the safety, resource utilization rate, and operation and maintenance efficiency of fly ash treatment. Summary of the Invention

[0003] The purpose of the present invention is to provide an intelligent monitoring system and method for a fly ash production line based on multi-source data to solve the problems raised in the prior art.

[0004] To solve the above technical problems, the present invention provides the following technical solution: An intelligent monitoring method for a fly ash production line based on multi-source data, the method comprising: Step S100: Calculate the initial composition score of fly ash through historical monitoring records, determine the initial grade of fly ash, and determine the characteristic records; Step S200: Calculate the environmental score, determine the set of characteristic process parameters of the processing procedure, and establish a classification model for the processing procedure; Step S300: Respectively construct prediction models for the finished product score and the environmental score, and establish a minimization objective function according to the label values of the characteristic records; Step S400: Based on the real-time composition of fly ash and the preparation target, combine the classification model to judge the label value, respectively construct the corresponding objective function, evaluate and sort the objective function values of the candidate processing procedures, and select the processing procedure with the smallest objective function value as the optimal processing procedure.

[0005] Further, step S100 includes: Step S101: In the fly ash production line monitoring system, obtain the type of finished product to be prepared from the fly ash, collect the initial component content of the fly ash, monitor the processing parameters of each processing procedure. After all processing procedures are completed, inspect the prepared finished product to generate a monitoring record. Step S102: Obtain historical monitoring records, summarize the historical monitoring records of preparing the same kind of finished product, collect the initial component content of the fly ash, and calculate the fly ash initial component score according to the following formula: ; where G represents the fly ash initial component score, H g represents the content of the g-th initial component, K g represents the weight of the g-th initial component, and h represents the total number of initial component types; Step S103: Summarize the fly ash initial component scores of all historical monitoring records, preset several initial levels, determine the fly ash initial component score interval corresponding to each initial level, and summarize the historical monitoring records of a certain initial level. Step S104: In the historical monitoring records of a certain initial level, collect the inspection result parameters of the finished product prepared in a certain historical monitoring record, perform normalization calculation on the inspection results, and calculate the finished product inspection score according to the following formula: ; where A represents the finished product inspection score, B a represents the a-th normalized inspection result parameter, C a represents the weight of the a-th inspection result, and b represents the total number of inspection results; Step S105: Preset a finished product inspection score threshold, set the historical monitoring records exceeding the finished product inspection score threshold as characteristic records, and summarize the characteristic records of a certain kind of finished product. By performing weighted scoring on the raw material components and the finished product quality, an objective and quantifiable data basis is achieved. Classifying according to the fly ash initial component score helps to optimize the process paths for raw materials of different grades. By comparing the characteristic records, it is possible to identify which processing parameter combinations are optimal under a specific raw material grade, providing a data support basis for the construction of an intelligent system for fly ash treatment, which is beneficial for automatic parameter adjustment and process self-optimization.

[0006] Further, step S200 includes: Step S201: Obtain the set of characteristic records of a certain kind of finished product, collect the environmental index parameters of a certain processing procedure in a certain characteristic record, and calculate the environmental score according to the following formula: ; Among them, S represents the environmental score, and D d represents the value of the d-th environmental index parameter, and E d represents the weight of the d-th environmental index parameter, and e represents the total number of environmental index parameters; Step S202: Preset an environmental score threshold, mark the feature records that exceed the environmental score threshold as abnormal records and set the label value to 1, and mark the feature records that do not exceed the environmental score threshold and set the label value to 0; Step S203: Obtain the process parameters in the abnormal records and construct a process parameter vector P = (P1, P2,..., Pn) T , where P1, P2,...Pn respectively represent the 1st, 2nd,...nth process parameters, combine a certain process parameter with the environmental score, calculate the Pearson correlation coefficient of the process parameter, set the Pearson correlation coefficient threshold, and set the process parameters that exceed the Pearson correlation coefficient threshold as feature parameters to obtain the set of feature process parameters for a certain processing operation; Step S204: Obtain all the feature records corresponding to a certain processing operation and establish a training data set , where Xi and Qi respectively represent the initial fly ash composition score and the set of feature process parameter values of the i-th feature record, Li represents the label value of the i-th feature record, N represents the total number of feature records corresponding to a certain processing operation, and establish a classification model f_L(X, Q) = L for a certain processing operation, where L ∈ {0, 1}, use X and Q in each training data set as input values and L as the output value, and train through a random forest model to obtain the classification model corresponding to a certain processing operation; Quickly identify processing records that may have environmental protection hazards through the environmental score and label mechanism, construct an abnormal database, and accurately screen out the variables that have the greatest impact on the environment from numerous process parameters with the help of Pearson correlation analysis to provide a basis for process optimization; The random forest model has non-linear modeling ability and strong generalization ability, can accurately identify under what conditions abnormalities are likely to occur, assist in early intervention, and adjust key process parameters in advance through the model prediction results to reduce pollutant emissions and improve the environmental protection level of the production line.

[0007] Furthermore, step S300 includes: Step S301: Aggregate the feature record sets of the same initial level, form a finished product inspection score data group with the fly ash initial composition score and the feature process parameter set of the feature record and the finished product inspection score, aggregate the finished product inspection score data groups of all feature records, establish a finished product inspection score prediction model f_A(X, Q)=A, use X and Q in each finished product inspection score data group as input values, and A as the output value, and train through a linear regression model to obtain the finished product inspection score prediction model; Step S302: Establish the objective function formula with a label value of 0 as: J1(Q)=[A’-f_A(X, Q)] 2 , where J1(Q) is the objective function value with a label value of 0, and A’ represents the preset target score value; Step S303: Obtain the feature records with a label value of 1 for a certain processing procedure, form an environmental score data group with the ash initial composition score and the feature process parameter set and the environmental score, aggregate the environmental score data groups of all feature records, establish an environmental score prediction model f_S(X, Q)=S, use X and Q in each environmental score data group as input values, and S as the output value, and train through a linear regression model to obtain the environmental score prediction model; Step S304: Establish the objective function formula with a label value of 1 as: J2(Q)=[A’-f_A(X, Q)] 2 +α×max[0, f_S(X, Q)-S’] 2 , where J2(Q) is the objective function value with a label value of 1, S’ represents the preset environmental score value, and α represents the weight value of the environmental score; No longer only pursue a single goal such as output and quality, but quantify the environmental protection risk as part of the model, supporting the design of an optimization path with the finished product score as the goal and the environmental protection score as the constraint; Compared with complex black-box models such as neural networks, linear regression provides higher interpretability, and it can clearly understand the positive and negative impacts of each feature process parameter on the finished product score and the environmental score; Support the production system to switch the objective function according to the real-time label value, adapt to different operation stages, and calculate the "optimal process parameter vector" by minimizing, which is conducive to giving automatic tuning suggestions for the closed-loop control system.

[0008] Further, step S400 includes: Step S401: Obtain the type of finished product to be prepared from the real-time fly ash, collect the real-time initial composition content of the fly ash, calculate the real-time initial composition score, and determine the real-time initial level of the fly ash; Step S402: Collect the real-time process parameters of a certain processing operation on fly ash, and input the real-time initial composition score of fly ash and the current set of characteristic process parameter values into the classification model of the processing operation. If the label value is 1, among multiple candidate processing operations, input the set of characteristic process parameter values into the objective function formula with a label value of 1 respectively, and calculate the objective function value of each candidate processing operation. If the label value is 0, among multiple candidate processing operations, input the set of characteristic process parameter values into the objective function formula with a label value of 0 respectively, and calculate the objective function value of each candidate processing operation; Step S403: Sort all the objective function values, and select the candidate processing operation corresponding to the minimum objective function value as the optimal operation; The system no longer depends on a fixed process, but intelligently recommends the optimal processing operation based on real-time raw materials and working conditions, which can cope with the actual problems of large fluctuations in fly ash raw materials and complex compositions, and achieve flexible manufacturing; Judge whether there is an environmental risk through the classification model, convert environmental compliance from post-treatment to pre-judgment, and use the penalty term in the objective function to effectively suppress the selection probability of potential pollution processes; Add double constraints on the finished product score and the environmental score to the objective function, make the optimization path more scientific and reasonable, and all candidate processes can be evaluated within the same algorithm framework to achieve automatic process scheduling control; The selection of traditional processes depends on the experience of operators, while this method relies on data modeling and algorithm judgment to improve consistency and response speed.

[0009] In order to better implement the above method, an intelligent monitoring system for fly ash production lines based on multi-source data is also proposed. The system includes a feature recording module, a classification model module, an objective function module, and a real-time sorting module; Feature recording module: Calculate the initial composition score of fly ash through historical monitoring records, determine the initial grade of fly ash, and determine the feature record; Classification model module: Calculate the environmental score, determine the set of characteristic process parameters of the processing operation, and establish a classification model of the processing operation; Objective function module: Respectively construct prediction models for the finished product score and the environmental score, and establish a minimized objective function according to the label value of the feature record; Real-time sorting module: Based on the real-time composition of fly ash and the preparation target, combine the classification model to judge the label value, respectively construct the corresponding objective function, evaluate and sort the objective function values of candidate processing operations, and select the operation with the minimum objective function value as the optimal processing operation.

[0010] Furthermore, the feature recording module includes an initial grade determination unit and a feature record determination unit: Determine the initial grade unit: In the fly ash production line monitoring system, obtain the type of finished product to be prepared from the fly ash, collect the initial component content of the fly ash, monitor the processing parameters of each processing step. After all processing steps are completed, inspect the prepared finished product to generate a monitoring record. Obtain the historical monitoring records, summarize the historical monitoring records of preparing the same finished product, collect the initial component content of the fly ash, calculate the initial component score of the fly ash, summarize the initial component scores of the fly ash in all historical monitoring records, preset several initial grades, and determine the fly ash initial component score interval corresponding to each initial grade; Determine the characteristic record unit: In the historical monitoring records of a certain initial grade, collect the test result parameters of the finished product prepared in a certain historical monitoring record, and perform normalization calculation on the test results, calculate the finished product test score, preset the finished product test score threshold, and set the historical monitoring records exceeding the finished product test score threshold as characteristic records.

[0011] Furthermore, the classification model module includes a determining abnormal record unit and a building classification model unit: Determine the abnormal record unit: Obtain the set of characteristic records of a certain finished product, collect the environmental index parameters of a certain processing step in a certain characteristic record, calculate the environmental score, preset the environmental score threshold, and set the characteristic records exceeding the environmental score threshold as abnormal records; Build the classification model unit: Obtain the process parameters in the abnormal records, construct a process parameter vector, combine a certain process parameter with the environmental score, calculate the Pearson correlation coefficient of the process parameter, set the Pearson correlation coefficient threshold, set the process parameters exceeding the Pearson correlation coefficient threshold as characteristic parameters, obtain the set of characteristic process parameters for a certain processing step, obtain all the characteristic records corresponding to a certain processing step, establish a training data set, and input the training data set into the random forest model to obtain the classification model corresponding to a certain processing step.

[0012] Furthermore, the objective function module includes a building target function unit with label value 0 and a building target function unit with label value 1: Build the target function unit with label value 0: Summarize the set of characteristic records of the same initial grade, form a finished product test score data group with the fly ash initial component score and the set of characteristic process parameters of the characteristic records and the finished product test score, summarize the finished product test score data groups of all characteristic records and input them into the linear regression model, establish a finished product test score prediction model, and establish the target function formula with label value 0; Establish a target function unit with a label value of 1: Obtain the feature records with a label value of 1 for a certain processing step, form an environmental score data group by combining the initial ash composition score and the set of characteristic process parameters with the environmental score, summarize the environmental score data groups of all feature records and input them into a linear regression model to establish an environmental score prediction model, and establish a target function formula with a label value of 1.

[0013] Furthermore, the real-time sorting module includes a unit for determining the real-time initial grade and a unit for selecting the optimal process: Unit for determining the real-time initial grade: Obtain the type of product to be prepared from the real-time fly ash, collect the real-time initial composition content of the fly ash, calculate the real-time initial composition score, and determine the real-time initial grade of the fly ash; Unit for selecting the optimal process: Collect the real-time process parameters of the fly ash when performing a certain processing step, input the real-time initial composition score of the fly ash and the current set of characteristic process parameter values into the classification model of the processing step. If the label value is 1, then among multiple candidate processing steps, respectively input the set of characteristic process parameter values into the target function formula with a label value of 1 to calculate the target function value of each candidate processing step. If the label value is 0, then among multiple candidate processing steps, respectively input the set of characteristic process parameter values into the target function formula with a label value of 0 to calculate the target function value of each candidate processing step. Sort all the target function values and select the candidate processing step corresponding to the minimum target function value as the optimal process.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention establishes a closed-loop process from raw material analysis → finished product detection → environmental assessment → process selection, and has multiple intelligent functions such as adaptive modeling, anomaly recognition, parameter tuning, and process scheduling, realizing the transformation of traditional fly ash treatment production lines to intelligent, green, and data-driven ones; Traditional optimization methods often only focus on a single objective such as output or quality. The present invention proposes that when the label is 0, only optimize the target function of the finished product score, and when the label is 1, optimize the multi-objective function of the finished product score + environmental constraints, having the ability to flexibly respond to different working conditions and automatically switch the optimization logic; Mark anomalies through environmental scores and thresholds, and use the classification model to predict whether the current parameter combination may be abnormal, realizing pre-judgment, real-time warning, and in-process intervention, and improving safety and stability. Description of the Drawings

[0015] Figure 1 It is a schematic flow chart of a method for intelligent monitoring of a fly ash production line based on multi-source data according to the present invention; Figure 2 It is a schematic structural diagram of a system for intelligent monitoring of a fly ash production line based on multi-source data according to the present invention. Detailed Embodiments

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] Please refer to Figure 1 and Figure 2 , the present invention provides a technical solution: an intelligent monitoring method for a fly ash production line based on multi-source data, and the method includes: Step S100: Calculate the initial component score of the fly ash, determine the initial grade of the fly ash, and determine the characteristic record through the historical monitoring record; Among them, step S100 includes: Step S101: In the fly ash production line monitoring system, obtain the type of product to be prepared from the fly ash, collect the initial component content of the fly ash, monitor the processing parameters of each processing procedure, and after all processing procedures are completed, detect the prepared product to generate a monitoring record; Step S102: Obtain the historical monitoring record, summarize the historical monitoring records of preparing the same product, collect the initial component content of the fly ash, and calculate the initial component score of the fly ash according to the following formula: ; Among them, G represents the initial component score of the fly ash, H g represents the content of the g-th initial component, K g represents the weight of the g-th initial component, and h represents the total number of types of initial components; Step S103: Summarize the initial component scores of all historical monitoring records, preset several initial grades, determine the fly ash initial component score interval corresponding to each initial grade, and summarize the historical monitoring records of a certain initial grade; Step S104: In the historical monitoring records of a certain initial grade, collect the detection result parameters of the product prepared by a certain historical monitoring record, perform normalization calculation on the detection results, and calculate the product detection score according to the following formula: ; Among them, A represents the product detection score, B a represents the a-th normalized detection result parameter, C a represents the weight of the a-th detection result, and b represents the total number of detection results; Step S105: Preset the finished product inspection score threshold, set the historical monitoring records exceeding the finished product inspection score threshold as characteristic records, and summarize the characteristic records of a certain finished product; For example, in the first monitoring record, the type of finished product is green building material bricks, the initial fly ash component content of cl is 4.2, Pb is 1.3, Zn is 0.9, and Hg is 0.01; the temperature in process 1 is 120 °C and the time is 30 min; the additive ratio in process 2 is 8%, and the pH value is controlled at 9.5; the heating temperature in process 3 is 200 °C and the wind speed is 5 m / s; the compressive strength is 0.88, the water absorption rate is 0.72, and the heavy metal residue in the tail gas is 0.65. Let the weight value of the compressive strength be 0.4, the weight value of the water absorption rate be 0.3, and the weight value of the heavy metal residue in the tail gas be 0.3; Let the weight value of cl be 0.3, the weight value of Pb be 0.3, the weight value of Zn be 0.2, and the weight value of Hg be 0.2, and calculate the initial fly ash component score to be 1.832; Calculate the finished product inspection score to be 0.763, and the preset finished product score threshold is 0.75, then the first monitoring record is a characteristic record.

[0018] Step S200: Calculate the environmental score, determine the set of characteristic process parameters for the processing procedure, and establish a classification model for the processing procedure; Among them, step S200 includes: Step S201: Obtain the set of characteristic records of a certain finished product, collect the environmental index parameters of a certain processing procedure in a certain characteristic record, and calculate the environmental score according to the following formula: ; Among them, S represents the environmental score, D d represents the value of the d-th environmental index parameter, E d represents the weight value of the d-th environmental index parameter, and e represents the total number of environmental index parameters; Step S202: Preset the environmental score threshold, set the characteristic records exceeding the environmental score threshold as abnormal records and mark them with a label value of 1, and mark the characteristic records not exceeding the environmental score threshold with a label value of 0; Step S203: Obtain the process parameters in the abnormal records, construct a process parameter vector P = (P1, P2,..., Pn) T , where P1, P2,... Pn respectively represent the 1st, 2nd,... nth process parameters, combine a certain process parameter with the environmental score, calculate the Pearson correlation coefficient of the process parameter, set the Pearson correlation coefficient threshold, and set the process parameters exceeding the Pearson correlation coefficient threshold as characteristic parameters to obtain the set of characteristic process parameters for a certain processing procedure; Step S204: Obtain all the characteristic records corresponding to a certain processing procedure and establish a training data set , where Xi and Qi respectively represent the initial fly ash composition score and the set of characteristic process parameter values of the i-th characteristic record, Li represents the label value of the i-th characteristic record, N represents the total number of characteristic records corresponding to a certain processing procedure, establish a classification model f_L(X, Q)=L corresponding to a certain processing procedure, where L∈{0, 1}, take X and Q in each training data set as input values and L as the output value, and train through a random forest model to obtain a classification model corresponding to a certain processing procedure; For example, in the first characteristic record, the heavy metal concentration in the discharge is 1.8 mg / L, the weight is 0.4, the ammonia nitrogen concentration in the tail gas is 36.0 ppm, the weight is 0.3, and the deviation degree of the pH value after solidification is 0.6, the weight is 0.3, and the calculated environmental score is 11.7; If the preset environmental score threshold is 10, then the first characteristic record is an abnormal record and the label value is 1.

[0019] Step S300: Construct prediction models for the finished product score and the environmental score respectively, and establish a minimization objective function according to the label values of the characteristic records; Among them, step S300 includes: Step S301: Aggregate the set of characteristic records of the same initial grade, form a finished product inspection score data group with the initial fly ash composition score and the set of characteristic process parameters of the characteristic records and the finished product inspection score, aggregate the finished product inspection score data groups of all characteristic records, establish a finished product inspection score prediction model f_A(X, Q)=A, take X and Q in each finished product inspection score data group as input values and A as the output value, and train through a linear regression model to obtain a finished product inspection score prediction model; Step S302: Establish the objective function formula with the label value of 0 as: J1(Q)=[A’ - f_A(X, Q)] 2 , where J1(Q) is the objective function value with the label value of 0, and A’ represents the preset target score value; Step S303: Obtain the characteristic records with the label value of 1 for a certain processing procedure, form an environmental score data group with the initial ash composition score and the set of characteristic process parameters and the environmental score, aggregate the environmental score data groups of all characteristic records, establish an environmental score prediction model f_S(X, Q)=S, take X and Q in each environmental score data group as input values and S as the output value, and train through a linear regression model to obtain an environmental score prediction model; Step S304: Establish the objective function formula with the label value of 1 as: J2(Q)=[A’ - f_A(X, Q)] 2 +α×max[0, f_S(X, Q)-S’]2 , where J2(Q) is the objective function value with a label value of 1, S’ represents a preset environmental score value, and α represents the weight of the environmental score.

[0020] Step S400: Based on the real-time composition of fly ash and the preparation target, combine the classification model to judge the label value, construct corresponding objective functions respectively, evaluate and sort the objective function values of candidate processing procedures, and select the procedure with the smallest objective function value as the optimal processing procedure.

[0021] Among them, step S400 includes: Step S401: Obtain the type of product to be prepared from the real-time fly ash, collect the real-time initial composition content of the fly ash, calculate the real-time initial composition score, and determine the real-time initial grade of the fly ash; Step S402: Collect the real-time process parameters of the fly ash when performing a certain processing procedure, input the real-time initial composition score of the fly ash and the current set of characteristic process parameter values into the classification model of the processing procedure. If the label value is 1, then among multiple candidate processing procedures, input the set of characteristic process parameter values into the objective function formula with a label value of 1 respectively, and calculate the objective function value of each candidate processing procedure. If the label value is 0, then among multiple candidate processing procedures, input the set of characteristic process parameter values into the objective function formula with a label value of 0 respectively, and calculate the objective function value of each candidate processing procedure; Step S403: Sort all the objective function values, and select the candidate processing procedure corresponding to the minimum objective function value as the optimal procedure; For example, the real-time initial composition content of the fly ash is cl = 3.8, Pb = 1.5, Zn = 1.1, Hg = 0.02. The calculated real-time initial composition score is 1.814, and the real-time initial grade of the fly ash is determined to be the second level; Obtain the current set of characteristic process parameter values, input them into the classification model, and get a label value of 1, and the objective function formula with a label value of 1 needs to be used; Suppose the set of characteristic process parameter values of the candidate processing procedure is as shown in Table 1: Table 1

[0022] The calculated objective function of process A is .1357, the objective function of process B is .6775, and the objective function of process C is .000676. Select process C as the optimal processing procedure for the current fly ash batch.

[0023] In order to better implement the above method, an intelligent monitoring system for fly ash production lines based on multi-source data is also proposed. The system includes a feature recording module, a classification model module, an objective function module, and a real-time sorting module; Feature recording module: Calculate the initial composition score of fly ash, determine the initial grade of fly ash, and determine feature records through historical monitoring records; Among them, the feature recording module includes an initial grade determination unit and a feature record determination unit: Initial grade determination unit: In the fly ash production line monitoring system, obtain the type of product to be prepared from fly ash, collect the initial composition content of fly ash, monitor the processing parameters of each processing step. After all processing steps are completed, detect the prepared product, generate a monitoring record, obtain historical monitoring records, summarize the historical monitoring records of preparing the same kind of product, collect the initial composition content of fly ash, calculate the initial composition score of fly ash, summarize the initial composition scores of all historical monitoring records, preset several initial grades, and determine the fly ash initial composition score interval corresponding to each initial grade; Feature record determination unit: In the historical monitoring records of a certain initial grade, collect the test result parameters of a certain historical monitoring record for the prepared product, perform normalization calculation on the test results, calculate the product test score, preset the product test score threshold, and set the historical monitoring records that exceed the product test score threshold as feature records.

[0024] Classification model module: Calculate the environmental score, determine the set of characteristic process parameters for the processing step, and establish a classification model for the processing step; Among them, the classification model module includes an abnormal record determination unit and a classification model establishment unit: Abnormal record determination unit: Obtain the set of feature records of a certain product, collect the environmental index parameters of a certain processing step in a certain feature record, calculate the environmental score, preset the environmental score threshold, and set the feature records that exceed the environmental score threshold as abnormal records; Classification model establishment unit: Obtain the process parameters in the abnormal records, construct a process parameter vector, combine a certain process parameter with the environmental score, calculate the Pearson correlation coefficient of the process parameter, set the Pearson correlation coefficient threshold, set the process parameters that exceed the Pearson correlation coefficient threshold as characteristic parameters, obtain the set of characteristic process parameters for a certain processing step, obtain all the feature records corresponding to a certain processing step, establish a training data set, and input the training data set into a random forest model to obtain the classification model corresponding to a certain processing step.

[0025] Objective function module: Construct prediction models for the product score and the environmental score respectively, and establish a minimization objective function according to the label values of the feature records; Among them, the objective function module includes a target function establishment unit with label value 0 and a target function establishment unit with label value 1: Establish a target function unit with a label value of 0: Summarize the set of feature records at the same initial level, form a finished product inspection score data group from the fly ash initial composition score and the set of characteristic process parameters of the feature records and the finished product inspection score, summarize the finished product inspection score data groups of all feature records and input them into the linear regression model, establish a finished product inspection score prediction model, and establish a target function formula with a label value of 0. Establish a target function unit with a label value of 1: Obtain the feature records with a label value of 1 for a certain processing operation, form an environmental score data group from the ash initial composition score and the set of characteristic process parameters and the environmental score, summarize the environmental score data groups of all feature records and input them into the linear regression model, establish an environmental score prediction model, and establish a target function formula with a label value of 1.

[0026] Real-time sorting module: Based on the real-time composition of fly ash and the preparation target, combine the classification model to judge the label value, respectively construct the corresponding target functions, evaluate and sort the target function values of the candidate processing operations, and select the processing operation with the minimum target function value as the optimal processing operation. Among them, the real-time sorting module includes a unit for determining the real-time initial level and a unit for selecting the optimal process: Unit for determining the real-time initial level: Obtain the type of finished product to be prepared from the real-time fly ash, collect the real-time initial composition content of the fly ash, calculate the real-time initial composition score, and determine the real-time initial level of the fly ash. Unit for selecting the optimal process: Collect the real-time process parameters of the fly ash for a certain processing operation, input the real-time initial composition score of the fly ash and the current set of characteristic process parameter values into the classification model of the processing operation. If the label value is 1, then among multiple candidate processing operations, respectively input the set of characteristic process parameter values into the target function formula with a label value of 1, calculate the target function value of each candidate processing operation. If the label value is 0, then among multiple candidate processing operations, respectively input the set of characteristic process parameter values into the target function formula with a label value of 0, calculate the target function value of each candidate processing operation, sort all the target function values, and select the candidate processing operation corresponding to the minimum target function value as the optimal process.

[0027] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to embrace all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

Claims

1. An intelligent monitoring method for fly ash production line based on multi-source data, characterized in that, The method includes: Step S100: Calculate the initial composition score of fly ash through historical monitoring records, determine the initial grade of fly ash, and determine the characteristic records; Step S200: Calculate the environmental score, determine the set of characteristic process parameters of the processing procedure, and establish a classification model of the processing procedure; Step S300: Construct prediction models for the finished product score and the environmental score respectively, and establish a minimization objective function according to the label values of the characteristic records; Step S400: Based on the real-time composition of fly ash and the preparation target, combine the classification model to judge the label value, construct the corresponding objective functions respectively, evaluate and sort the objective function values of the candidate processing procedures, and select the processing procedure with the smallest objective function value as the optimal processing procedure.

2. The intelligent monitoring method for fly ash production line based on multi-source data according to claim 1, characterized in that The said step S100 includes the following steps: Step S101: In the fly ash production line monitoring system, obtain the type of finished product to be prepared from fly ash, collect the initial composition content of fly ash, monitor the processing parameters of each processing procedure, and after all processing procedures are completed, detect the prepared finished product to generate a monitoring record; Step S102: Obtain the historical monitoring records, summarize the historical monitoring records of preparing the same kind of finished product, collect the initial composition content of fly ash, and calculate the initial composition score of fly ash according to the following formula: ； Among them, G represents the initial component score of fly ash, and H g represents the content of the g-th initial component, and K g represents the weight of the g-th initial component, and h represents the total number of types of initial components; Step S103: Summarize the initial composition scores of fly ash in all historical monitoring records, preset several initial grades, determine the fly ash initial composition score interval corresponding to each initial grade, and summarize the historical monitoring records of a certain initial grade; Step S104: In the historical monitoring records of a certain initial grade, collect the detection result parameters of the finished product prepared by a certain historical monitoring record, perform normalization calculation on the detection results, and calculate the finished product detection score according to the following formula: ； Among them, A represents the finished product inspection score, and B a represents the a-th normalized inspection result parameter, and C a represents the weight of the a-th inspection result, and b represents the total number of inspection results; Step S105: Preset the finished product detection score threshold, set the historical monitoring records exceeding the finished product detection score threshold as characteristic records, and summarize the characteristic records of a certain kind of finished product.

3. The intelligent monitoring method for fly ash production line based on multi-source data according to claim 2, wherein, The said step S200 includes the following steps: Step S$201: Obtain the set of characteristic records of a certain kind of finished product, collect the environmental index parameters of a certain processing procedure in a certain characteristic record, and calculate the environmental score according to the following formula: ； Among them, S represents the environmental score, and D d represents the value of the d-th environmental index parameter, and E d represents the weight of the d-th environmental index parameter, and e represents the total number of environmental index parameters; Step S202: Preset the environmental score threshold, set the characteristic records exceeding the environmental score threshold as abnormal records and mark them with a label value of 1, and mark the characteristic records not exceeding the environmental score threshold with a label value of 0; Step S203: Obtain the process parameters in the exception record, and construct a process parameter vector P = (P1, P2,..., Pn), where P1, P2,..., Pn represent the 1st, 2nd,..., nth process parameters respectively. Combine a certain process parameter with the environmental score, calculate the Pearson correlation coefficient of the process parameter, set a Pearson correlation coefficient threshold, and set the process parameters exceeding the Pearson correlation coefficient threshold as characteristic parameters to obtain a set of characteristic process parameters for a certain processing operation. T , where P1, P2,..., Pn represent the 1st, 2nd,..., nth process parameters respectively. Combine a certain process parameter with the environmental score, calculate the Pearson correlation coefficient of the process parameter, set a Pearson correlation coefficient threshold, and set the process parameters exceeding the Pearson correlation coefficient threshold as characteristic parameters to obtain a set of characteristic process parameters for a certain processing operation. Step S204: Obtain all the feature records corresponding to a certain processing operation, and establish a training data set , where Xi and Qi respectively represent the fly ash initial composition score and the set of characteristic process parameter values of the i-th feature record, Li represents the label value of the i-th feature record, N represents the total number of feature records corresponding to a certain processing operation, establish a classification model f_L(X, Q)=L corresponding to a certain processing operation, where L∈{0, 1}, use X and Q in each training data set as input values and L as the output value, and train through a random forest model to obtain a classification model corresponding to a certain processing operation.

4. The intelligent monitoring method for fly ash production line based on multi-source data according to claim 3, wherein, The said step S300 includes the following steps: Step S301: Summarize the set of characteristic records of the same initial grade, form a finished product detection score data group with the fly ash initial composition score and the set of characteristic process parameters of the characteristic records and the finished product detection score, summarize the finished product detection score data groups of all characteristic records, establish a finished product detection score prediction model f_A(X,Q)=A, take X and Q in each finished product detection score data group as input values, and A as the output value, and train through a linear regression model to obtain the finished product detection score prediction model; Step S302: Establish the objective function formula with a label value of 0 as: J1(Q) = [A’ - f_A(X, Q)] 2 , where J1(Q) is the objective function value with a label value of 0, and A’ represents a preset target score value; Step S303: Obtain the feature records with the label value of 1 for a certain processing step, form an environmental score data group by combining the initial ash composition score, the set of characteristic process parameters, and the environmental score, summarize the environmental score data groups of all feature records, and establish an environmental score prediction model f_S(X,Q)=S. Use X and Q in each environmental score data group as input values and S as the output value, and train through a linear regression model to obtain the environmental score prediction model; Step S304: Establish the objective function formula with a label value of 1 as: J2(Q) = [A’ - f_A(X,Q)] 2 + α × max[0, f_S(X,Q) - S’] 2 , where J2(Q) is the objective function value with a label value of 1, S’ represents the preset environmental score value, and α represents the weight of the environmental score.

5. The intelligent monitoring method for fly ash production line based on multi-source data according to claim 4, wherein The step S400 includes the following steps: Step S401: Obtain the type of product to be prepared from the real-time fly ash, collect the real-time initial composition content of the fly ash, calculate the real-time initial composition score, and determine the real-time initial grade of the fly ash; Step S402: Collect the real-time process parameters of the fly ash during a certain processing step. Input the real-time initial composition score of the fly ash and the current set of characteristic process parameter values into the classification model of the processing step. If the label value is 1, then among multiple candidate processing steps, input the set of characteristic process parameter values into the objective function formula with a label value of 1 respectively to calculate the objective function value of each candidate processing step. If the label value is 0, then among multiple candidate processing steps, input the set of characteristic process parameter values into the objective function formula with a label value of 0 respectively to calculate the objective function value of each candidate processing step; Step S403: Sort all the objective function values, and select the candidate processing step corresponding to the minimum objective function value as the optimal step.

6. An intelligent monitoring system for a fly ash production line based on multi-source data, which is used to implement the intelligent monitoring method for a fly ash production line based on multi-source data described in any one of claims 1-5, characterized in that, The system includes a feature record module, a classification model module, an objective function module, and a real-time sorting module; The feature record module: Calculate the initial ash composition score through historical monitoring records, determine the initial grade of the fly ash, and determine the feature records; The classification model module: Calculate the environmental score, determine the set of characteristic process parameters of the processing step, and establish a classification model of the processing step; The objective function module: Respectively construct prediction models for the finished product score and the environmental score, and establish a minimization objective function according to the label value of the feature record; The real-time sorting module: Based on the real-time composition of the fly ash and the preparation target, combine the classification model to judge the label value, respectively construct the corresponding objective function, evaluate and sort the objective function values of the candidate processing steps, and select the step with the minimum objective function value as the optimal processing step.

7. The intelligent monitoring system for fly ash production line based on multi-source data according to claim 6, characterized in that, The feature record module includes an initial grade determination unit and a feature record determination unit: The initial grade determination unit: In the fly ash production line monitoring system, obtain the type of product to be prepared from the fly ash, collect the initial composition content of the fly ash, monitor the processing parameters of each processing step. After all processing steps are completed, detect the prepared finished product to generate a monitoring record. Obtain the historical monitoring records, summarize the historical monitoring records of preparing the same type of finished product, collect the initial composition content of the fly ash, calculate the initial ash composition score, summarize the initial ash composition scores of all historical monitoring records, preset several initial grades, and determine the fly ash initial composition score interval corresponding to each initial grade; The said determining feature recording unit: In the historical monitoring records of a certain initial level, collect the test result parameters of a certain historical monitoring record for the completed finished product, perform normalization calculation on the test results, calculate the finished product test score, preset the finished product test score threshold, and set the historical monitoring records exceeding the finished product test score threshold as feature records.

8. The intelligent monitoring system for fly ash production line based on multi-source data according to claim 6, characterized in that, The said classification model module includes a determining abnormal record unit and a building classification model unit: The said determining abnormal record unit: Obtain the set of feature records of a certain finished product, collect the environmental index parameters of a certain processing procedure in a certain feature record, calculate the environmental score, preset the environmental score threshold, and set the feature records exceeding the environmental score threshold as abnormal records; The said building classification model unit: Obtain the process parameters in the abnormal records, construct a process parameter vector, combine a certain process parameter with the environmental score, calculate the Pearson correlation coefficient of the said process parameter, set the Pearson correlation coefficient threshold, and set the process parameters exceeding the Pearson correlation coefficient threshold as feature parameters to obtain the set of feature process parameters of a certain processing procedure, obtain all the feature records corresponding to a certain processing procedure, build a training data set, input the training data set into the random forest model, and obtain the classification model corresponding to a certain processing procedure.

9. The intelligent monitoring system for fly ash production line based on multi-source data according to claim 6, wherein The said objective function module includes a building objective function unit with label value 0 and a building objective function unit with label value 1: The said building objective function unit with label value 0: Summarize the set of feature records of the same initial level, form a finished product test score data group with the fly ash initial composition score and the set of feature process parameters of the feature record and the finished product test score, summarize the finished product test score data groups of all feature records and input them into the linear regression model, build a finished product test score prediction model, and build an objective function formula with label value 0; The said building objective function unit with label value 1: Obtain the feature records with label value 1 of a certain processing procedure, form an environmental score data group with the ash initial composition score and the set of feature process parameters and the environmental score, summarize the environmental score data groups of all feature records and input them into the linear regression model, build an environmental score prediction model, and build an objective function formula with label value 1.

10. The intelligent monitoring system for fly ash production line based on multi-source data according to claim 6, characterized in that, The said real-time sorting module includes a determining real-time initial level unit and a selecting optimal procedure unit: The said determining real-time initial level unit: Obtain the type of finished product to be prepared from the real-time fly ash, collect the real-time initial composition content of the fly ash, calculate the real-time initial composition score, and determine the real-time initial level of the said fly ash; The optimal process unit selection: Collect the real-time process parameters of fly ash performing a certain processing operation, input the real-time initial composition score of fly ash and the set of current characteristic process parameter values into the classification model of the processing operation. If the label value is 1, among multiple candidate processing operations, input the set of characteristic process parameter values into the objective function formula with a label value of 1 respectively, and calculate the objective function value of each candidate processing operation. If the label value is 0, among multiple candidate processing operations, input the set of characteristic process parameter values into the objective function formula with a label value of 0 respectively, and calculate the objective function value of each candidate processing operation. Sort all the objective function values, and select the candidate processing operation corresponding to the minimum objective function value as the optimal process.

Images