Intelligent detection system and method for fly ash recycling products

Abstract

The invention discloses an intelligent detection system and method for fly ash recycling products, and relates to the technical field of data analysis, the system comprises a data acquisition module, a data processing module, a model classification module and a representative value calculation module; the data acquisition module is used for deploying a sensor to acquire fly ash recycling product characteristic data and storing the fly ash recycling product characteristic data into a database; the data processing module is used for identifying and processing missing values and abnormal values in the data and improving the data quality by using a time sequence model; the model classification module is used for classifying time sequence models of different deployment points; and the representative value calculation module is used for calculating representative values of product characteristics at different deployment points. The invention also provides a method for realizing the system. The method can effectively improve the condition that the detection data is lack of representativeness due to the fact that the product sampling sample is difficult to fully represent the whole sample in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to an intelligent detection system and method for fly ash resource products. Background Art

[0002] As the world pays more and more attention to environmental protection, countries have introduced strict environmental policies and regulations to control the discharge and disposal of waste. For fly ash, a hazardous waste containing heavy metals and organic pollutants, its resource utilization process must be strictly regulated.

[0003] In addition, fly ash contains many valuable components such as silicon, aluminum, and calcium. Testing fly ash resource products can optimize the use of these components, accurately evaluate product quality, and make fly ash more efficiently used in building materials, road materials, etc. Testing the content of harmful substances in products, such as heavy metals and radioactive substances, can also protect the public from potential health hazards.

[0004] During the detection of fly ash resource products, since the fly ash may have uneven composition during the production and storage process, it is difficult to fully represent the entire sample during sampling, which will lead to deviations in the test results; in addition, when sampling, the selected part is difficult to represent the entire area, and the influence of other components is ignored, which will also make the test data lack representativeness. Summary of the invention

[0005] The object of the present invention is to provide an intelligent detection system and method for fly ash resource products to solve the problems raised in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solutions: an intelligent detection system for fly ash resource products, the system comprising: a data acquisition module, a data processing module, a model classification module and a representative value calculation module; The data acquisition module is used to deploy sensors to collect fly ash resource product characteristic data and store it in the database; the data processing module is used to identify and process missing values ​​and abnormal values ​​in the data, and use the time series model to improve the data quality; the model classification module is used to classify the time series models of different deployment points; the representative value calculation module is used to calculate the representative values ​​of product characteristics at different deployment points; The output end of the data acquisition module is connected to the input end of the data processing module; the output end of the data processing module is connected to the input end of the model classification module; the output end of the model classification module is connected to the input end of the representative value calculation module.

[0007] The data acquisition module includes: a sensor deployment unit, a data acquisition unit, and a database unit; The sensor deployment unit is used to deploy IoT sensors at multiple locations; the data acquisition unit is used to collect fly ash resource product characteristic data at set time intervals; the database unit is used to build a database to store data and provide raw data for subsequent analysis; The output end of the sensor deployment unit is connected to the input end of the data acquisition unit; the output end of the data acquisition unit is connected to the input end of the database unit; the output end of the database unit is connected to the input end of the data processing module.

[0008] The data processing module includes: a preliminary processing unit, a time series model unit and a data processing unit; The preliminary processing unit is used to identify and preliminarily process missing values ​​and abnormal values ​​in the data; the time series model unit is used to construct a time series model based on the preliminarily processed data; the data processing unit is used to further interpolate and replace data based on the time series model; The output end of the preliminary processing unit is connected to the input end of the timing model unit; the output end of the timing model unit is connected to the input end of the data processing unit; the output end of the data processing unit is connected to the input end of the model classification module.

[0009] The model classification module includes: a preliminary classification unit, a cluster number setting unit and a model classification unit; The preliminary classification unit is used to preliminarily classify the model according to the order of the autoregressive term and the moving average term; the cluster number setting unit is used to set the cluster number of the K-means clustering model; the model classification unit is used to classify the model parameter vector according to the K-means clustering, and classify the models of different deployment points into different categories; The output end of the preliminary classification unit is connected to the input end of the cluster number setting unit; the output end of the cluster number setting unit is connected to the input end of the model classification unit; the output end of the model classification unit is connected to the input end of the representative value calculation unit.

[0010] The representative value calculation module includes: a representative value calculation unit; The representative value calculation unit is used to calculate the representative values ​​of product characteristics of different deployment points according to the model classification results, and perform weighted calculation by comprehensively considering the characteristic average values ​​of the deployment points to which models of different categories belong and the number of models.

[0011] An intelligent detection method for fly ash resource products, the method comprising the following steps: Step 1: Deploy IoT sensors, set the collection time interval, and build a database to collect fly ash resource product characteristic data at different times at each deployment point; Step 2: Linearly interpolate missing values, identify and process outliers, and build an ARMA model to interpolate missing values ​​and replace outliers; Step 3: Preliminarily classify the ARMA models by order, and then use K-means clustering to classify the model parameter vectors, so as to classify the models into corresponding categories; Step 4: Calculate the representative values ​​of product characteristics at different deployment points based on the average characteristics of the deployment points to which models of different categories belong and the number of models.

[0012] In step 1, according to the characteristics of the fly ash resource product to be tested [C1, C2, …, C u ], select the corresponding IoT sensor for deployment; select multiple IoT sensor deployment points: [D1,D2,…,D v ]; Where u is a positive integer, indicating the number of fly ash resource product characteristics; C u represents the characteristics of the u-th fly ash resource product; v is a positive integer, indicating the number of IoT sensor deployment points; D v represents the vth IoT sensor deployment point; Set the sensor data collection time interval to T; build a database to collect all the data collected during the period 0 to w·T; for the deployment point Da, the corresponding data is expressed as: [{t0:V(C1),V(C2),…,V(C u )},{t1:V(C1),V(C2),…,V(C u )},…,{t w :V(C1),V(C2),…,V(C u )}]; collect data collected from all deployment points; Where w is a positive integer, indicating the number of times data is collected; a is a positive integer, a∈{1,2,…,v}, D a represents the ath IoT sensor deployment point; t w represents the time w·T; V(C u ) represents the value of the characteristic of the u-th fly ash resource product.

[0013] In step 2, missing values ​​and outliers in the database are identified and processed; For deployment point D a Collected product characteristics C b Data, t i The missing values ​​at the moment are preliminarily processed using linear interpolation: V(C b ) ti insert =V(C b ) t(i-1) +(V(Cb ) t(i+1) -V(C b ) t(i-1) ) / (t (i+1) -t (i-1) )·(t i -t (i-1) ); Where i is a positive integer, representing the time sequence, i∈{1,2,…,w-1}; b is a positive integer, b∈{1,2,…,u}, C b represents the characteristics of the bth fly ash resource product; V(C b ) ti insert Indicates t i The interpolation value of the fly ash resource product characteristics at time b; V(C b ) t(i+1) Indicates t (i+1) The collected values ​​of the characteristics of the fly ash resource product at time b; Process missing values ​​in the data collected at each deployment point; Use the Z-score method to identify outliers; for deployment point D a Collected product characteristics C b , t j The data V(C b ) tj :Z(C b ) tj =(V(C b ) tj -μ(C b )) / σ(C b ); Set the threshold K, when |Z(C b ) tj When |>K, the value is considered an outlier; For outliers, use the average value of neighboring values ​​to replace the outliers for preliminary processing: V(C b ) tj instead =(V(C b ) t(j-1) +V(C b ) t(j+1) ) / 2; Where j is a positive integer, representing a time sequence, j∈{1,2,…,w-1}; V(C b ) tj instead Indicates t j The replacement value of the fly ash resource product characteristics at time b; V(C b ) t(j+1) Indicates t (j+1) The collection value of the characteristics of the fly ash resource product at the bth time; μ(Cb ) is product characteristic C b The mean value of σ(C b ) is product characteristic C b The standard deviation of Identify and process outliers in the data collected at each deployment point; For deployment point D a Collected product characteristics C b , construct an ARMA time series model: Input is [V(C b ) t0 ,V(C b ) t1 ,…,V(C b ) tw ]; Use ACF and PACF plots to determine AR order p and MA order q; Establish an ARMA model: V(C b ) ts =ϕ1V(C b ) t(s−1) +ϕ2V(C b ) t(s−2) +…+ϕ p V(C b ) t(s−p) +θ1ε t(s−1) +θ2ε t(s−2) +…+θ q ε t(s−q) +ε ts ; Fit the ARMA model according to the input and estimate the parameters by the maximum likelihood estimation method; In the trained model, use the model to predict the value of the missing position to interpolate the missing values; For the identified outliers, use the model to predict their normal values ​​for replacement; Where s is a positive integer, representing the time series, s ≥ p, s ≥ q; p represents the order of the autoregressive term; q represents the order of the moving average term; ϕ1, ϕ2, …, ϕ p Represents the parameters of the autoregressive part AR; θ1, θ2, …, θ q Represents the parameters of the moving average MA part; ε ts represents random error; Different ARMA time series model formulas are obtained by constructing time series models for different product characteristics at different deployment points.

[0014] In step 3, for different deployment point product features C b , there are v different ARMA time series model formulas; the v formulas are preliminarily classified according to the order p of the autoregressive term and the order q of the moving average term: formulas with the same p and formulas with the same q are classified into one category; For the formulas after preliminary classification, reclassify them: for the formulas with the same p and q, the parameters ϕ1, ϕ2, …, ϕ p and θ1,θ2,…,θ q Construct the vector [ϕ1,ϕ2,…,ϕ p ,θ1,θ2,…,θ q ]; Set the number of clusters K1, and use K-means clustering to cluster the same preliminary classification of different deployment point product features C b The corresponding ARMA time series model formula parameter vector [ϕ1,ϕ2,…,ϕ p ,θ1,θ2,…,θ q ] is divided into K1 classes; According to the different results of the preliminary classification, the formula parameters are divided into K1 categories: [L1, L2, ..., L K1 ], K2 classes: [L K1+1 ,L K1+2 ,…,L K1+K2 ], ..., K r Class: [L K1+K2+…+K(r-1)+1 ,L K1+K2+…+K(r-1)+2 ,…,L K1+K2+…+K(r-1)+Kr ]; formula parameters correspond to product features at different deployment points C b The v ARMA time series models are also divided into corresponding categories; Among them, r is a positive integer, which indicates the number of types of preliminary classification; L K1+K2+…+K(r-1)+Kr Indicates K1+K2+…+K r categories; The ARMA time series models corresponding to different product characteristics at different deployment points are classified.

[0015] In step S4, for the product characteristics C at different deployment points at a certain moment b , calculate the representative value of this characteristic: R=(V(C b ) L1 mean ×H L1 / H K1 +V(C b ) L2 mean ×H L2 / H K1 +…+V(C b ) LK1 mean ×H LK1 / H K1 )×K1 / (K1+K2+…+K r )+(V(C b ) L(K1+1) mean ×H L(K1+1) / HK2 +V(C b ) L(K1+2) mean ×H L(K1+2) / H K2 +…+V(C b ) L(K1+K2) mean ×H L(K1+K2) / H K2 )×K2 / (K1+K2+…+K r )+…+(V(C b ) L(K1+K2+…+K(r-1)+1) mean ×H L(K1+K2+…+K(r-1)+1) / H Kr +V(C b ) L(K1+K2+…+K(r-1)+2) mean ×H L(K1+K2+…+K(r-1)+2) / H Kr +…+V(C b ) L(K1+K2+…+Kr) mean ×H L(K1+K2+…+Kr) / H Kr )×K r / (K1+K2+…+K r ); Among them, V(C b ) L(K1+K2+…+K(r-1)+2) mean Indicates L K1+K2+…+K(r-1)+Kr Category corresponds to the product characteristics of all deployment points to which the time series model belongs C b Average value; H L(K1+K2+…+Kr) Indicates L K1+K2+…+K(r-1)+Kr The categories correspond to the number of all time series models; H Kr K r The categories correspond to the number of all time series models; Calculate representative values ​​for all product features at different deployment points.

[0016] A preliminary classification is performed based on the order p of the autoregressive term and the order q of the moving average term. Different p and q values ​​determine the structure and characteristics of the time series model. Models with the same p and q have similar dynamic behaviors and prediction capabilities to a certain extent. By classifying them into one category, the models can be preliminarily grouped for subsequent more refined classification. Further classification can analyze the distribution and change patterns of product characteristics at different deployment points, and also provide a more accurate classification basis for the subsequent calculation of representative values.

[0017] The representative value of each product characteristic is calculated by weighted summation. First, for each category, the average product characteristic value of the deployment point corresponding to the category is multiplied by the ratio of the number of models in the category to the total number of models in the category, and then multiplied by the weight corresponding to the category. Finally, the weighted results of all categories are added together to obtain the representative value of the product characteristic. Comprehensively considering the contribution of different category models to the representative value, the representative value can more accurately reflect the overall characteristics of the fly ash resource product.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention processes missing values ​​and outliers through a variety of methods, thereby improving the quality and reliability of data and reducing erroneous analysis and decision-making caused by data problems; the present invention classifies the ARMA time series models of different deployment points, which can better understand the differences and similarities in product characteristics at different locations, and provide a basis for more accurate characteristic evaluation; the ARMA model can capture the dynamic change characteristics of time series data, and compared with the static analysis method, it can reflect the change trend of fly ash characteristics in real time and provide a more timely decision-making basis; the present invention calculates representative values ​​by integrating the results of different categories of models, which can more accurately reflect the overall characteristics of fly ash resource products and provide more valuable information for product quality evaluation and resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of a process of an intelligent detection system for fly ash resource products of the present invention; Figure 2 The present invention is a schematic diagram of the steps of an intelligent detection method for fly ash resource products. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, an intelligent detection system for fly ash resource products, the system comprising: a data acquisition module, a data processing module, a model classification module and a representative value calculation module; The data acquisition module is used to deploy sensors to collect fly ash resource product characteristic data and store it in the database; the data processing module is used to identify and process missing values ​​and abnormal values ​​in the data, and use the time series model to improve the data quality; the model classification module is used to classify the time series models of different deployment points; the representative value calculation module is used to calculate the representative values ​​of product characteristics at different deployment points; The output end of the data acquisition module is connected to the input end of the data processing module; the output end of the data processing module is connected to the input end of the model classification module; the output end of the model classification module is connected to the input end of the representative value calculation module.

[0022] The data acquisition module includes: a sensor deployment unit, a data acquisition unit, and a database unit; The sensor deployment unit is used to deploy IoT sensors at multiple locations; the data acquisition unit is used to collect fly ash resource product characteristic data at set time intervals; the database unit is used to build a database to store data and provide raw data for subsequent analysis; The output end of the sensor deployment unit is connected to the input end of the data acquisition unit; the output end of the data acquisition unit is connected to the input end of the database unit; the output end of the database unit is connected to the input end of the data processing module.

[0023] The data processing module includes: a preliminary processing unit, a time series model unit and a data processing unit; The preliminary processing unit is used to identify and preliminarily process missing values ​​and abnormal values ​​in the data; the time series model unit is used to construct a time series model based on the preliminarily processed data; the data processing unit is used to further interpolate and replace data based on the time series model; The output end of the preliminary processing unit is connected to the input end of the timing model unit; the output end of the timing model unit is connected to the input end of the data processing unit; the output end of the data processing unit is connected to the input end of the model classification module.

[0024] The model classification module includes: a preliminary classification unit, a cluster number setting unit and a model classification unit; The preliminary classification unit is used to preliminarily classify the model according to the order of the autoregressive term and the moving average term; the cluster number setting unit is used to set the cluster number of the K-means clustering model; the model classification unit is used to classify the model parameter vector according to the K-means clustering, and classify the models of different deployment points into different categories; The output end of the preliminary classification unit is connected to the input end of the cluster number setting unit; the output end of the cluster number setting unit is connected to the input end of the model classification unit; the output end of the model classification unit is connected to the input end of the representative value calculation unit.

[0025] The representative value calculation module includes: a representative value calculation unit; The representative value calculation unit is used to calculate the representative values ​​of product characteristics of different deployment points according to the model classification results, and perform weighted calculation by comprehensively considering the characteristic average values ​​of the deployment points to which models of different categories belong and the number of models.

[0026] An intelligent detection method for fly ash resource products, the method comprising the following steps: Step 1: Deploy IoT sensors, set the collection time interval, and build a database to collect fly ash resource product characteristic data at different times at each deployment point; Step 2: Linearly interpolate missing values, identify and process outliers, and build an ARMA model to interpolate missing values ​​and replace outliers; Step 3: Preliminarily classify the ARMA models by order, and then use K-means clustering to classify the model parameter vectors, so as to classify the models into corresponding categories; Step 4: Calculate the representative values ​​of product characteristics at different deployment points based on the average characteristics of the deployment points to which models of different categories belong and the number of models.

[0027] In step 1, according to the characteristics of the fly ash resource product to be tested [C1, C2, …, C u ], select the corresponding IoT sensor for deployment; select multiple IoT sensor deployment points: [D1,D2,…,D v ]; Where u is a positive integer, indicating the number of fly ash resource product characteristics; C u represents the characteristics of the u-th fly ash resource product; v is a positive integer, indicating the number of IoT sensor deployment points; D v represents the vth IoT sensor deployment point; Set the sensor data collection time interval to T; build a database to collect all the data collected during the period 0 to w·T; for the deployment point Da, the corresponding data is expressed as: [{t0:V(C1),V(C2),…,V(C u )},{t1:V(C1),V(C2),…,V(C u )},…,{t w :V(C1),V(C2),…,V(C u )}]; collect data collected from all deployment points; Where w is a positive integer, indicating the number of times data is collected; a is a positive integer, a∈{1,2,…,v}, D a represents the ath IoT sensor deployment point; t w represents the time w·T; V(C u ) represents the value of the characteristic of the u-th fly ash resource product.

[0028] In step 2, missing values ​​and outliers in the database are identified and processed; For deployment point D a Collected product characteristics C b Data, t i The missing values ​​at the moment are preliminarily processed using linear interpolation: V(C b ) ti insert =V(C b ) t(i-1) +(V(C b ) t(i+1) -V(C b ) t(i-1) ) / (t (i+1) -t (i-1) )·(t i -t (i-1) ); Where i is a positive integer, representing the time sequence, i∈{1,2,…,w-1}; b is a positive integer, b∈{1,2,…,u}, C b represents the characteristics of the bth fly ash resource product; V(C b ) ti insert Indicates t i The interpolation value of the fly ash resource product characteristics at time b; V(C b ) t(i+1) Indicates t (i+1) The collected values ​​of the characteristics of the fly ash resource product at time b; Process missing values ​​in the data collected at each deployment point; Use the Z-score method to identify outliers; for deployment point D a Collected product characteristics C b , t j The data V(C b ) tj :Z(C b ) tj =(V(C b ) tj -μ(C b )) / σ(C b ); Set the threshold K, when |Z(C b )tj When |>K, the value is considered an outlier; For outliers, use the average value of neighboring values ​​to replace the outliers for preliminary processing: V(C b ) tj instead =(V(C b ) t(j-1) +V(C b ) t(j+1) ) / 2; Where j is a positive integer, representing a time sequence, j∈{1,2,…,w-1}; V(C b ) tj instead Indicates t j The replacement value of the fly ash resource product characteristics at time b; V(C b ) t(j+1) Indicates t (j+1) The collection value of the characteristics of the fly ash resource product at the bth time; μ(C b ) is product characteristic C b The mean value of σ(C b ) is product characteristic C b The standard deviation of Identify and process outliers in the data collected at each deployment point; For deployment point D a Collected product characteristics C b , construct an ARMA time series model: Input is [V(C b ) t0 ,V(C b ) t1 ,…,V(C b ) tw ]; Use ACF and PACF plots to determine AR order p and MA order q; Establish an ARMA model: V(C b ) ts =ϕ1V(C b ) t(s−1) +ϕ2V(C b ) t(s−2) +…+ϕ p V(C b ) t(s−p) +θ1ε t(s−1) +θ2ε t(s−2) +…+θ q ε t(s−q) +ε ts ; Fit the ARMA model according to the input and estimate the parameters by the maximum likelihood estimation method; In the trained model, use the model to predict the value of the missing position to interpolate the missing values; For the identified outliers, use the model to predict their normal values ​​for replacement; Where s is a positive integer, representing the time series, s ≥ p, s ≥ q; p represents the order of the autoregressive term; q represents the order of the moving average term; ϕ1, ϕ2, …, ϕ p Represents the parameters of the autoregressive part AR; θ1, θ2, …, θ q Represents the parameters of the moving average MA part; ε ts represents random error; Different ARMA time series model formulas are obtained by constructing time series models for different product characteristics at different deployment points.

[0029] In step 3, for different deployment point product features C b , there are v different ARMA time series model formulas; the v formulas are preliminarily classified according to the order p of the autoregressive term and the order q of the moving average term: formulas with the same p and formulas with the same q are classified into one category; For the formulas after preliminary classification, reclassify them: for the formulas with the same p and q, the parameters ϕ1, ϕ2, …, ϕ p and θ1,θ2,…,θ q Construct the vector [ϕ1,ϕ2,…,ϕ p ,θ1,θ2,…,θ q ]; Set the number of clusters K1, and use K-means clustering to cluster the same preliminary classification of different deployment point product features C b The corresponding ARMA time series model formula parameter vector [ϕ1,ϕ2,…,ϕ p ,θ1,θ2,…,θ q ] is divided into K1 classes; According to the different results of the preliminary classification, the formula parameters are divided into K1 categories: [L1, L2, ..., L K1 ], K2 classes: [L K1+1 ,L K1+2 ,…,L K1+K2 ], ..., K r Class: [L K1+K2+…+K(r-1)+1 ,L K1+K2+…+K(r-1)+2 ,…,L K1+K2+…+K(r-1)+Kr ]; formula parameters correspond to product features at different deployment points C b The v ARMA time series models are also divided into corresponding categories; Among them, r is a positive integer, which indicates the number of types of preliminary classification; L K1+K2+…+K(r-1)+Kr Indicates K1+K2+…+K r categories; The ARMA time series models corresponding to different product characteristics at different deployment points are classified.

[0030] In step S4, for the product characteristics C at different deployment points at a certain moment b , calculate the representative value of this characteristic: R=(V(C b ) L1 mean ×H L1 / H K1 +V(C b ) L2 mean ×H L2 / H K1 +…+V(C b ) LK1 mean ×H LK1 / H K1 )×K1 / (K1+K2+…+K r )+(V(C b ) L(K1+1) mean ×H L(K1+1) / H K2 +V(C b ) L(K1+2) mean ×H L(K1+2) / H K2 +…+V(C b ) L(K1+K2) mean ×H L(K1+K2) / H K2 )×K2 / (K1+K2+…+K r )+…+(V(C b ) L(K1+K2+…+K(r-1)+1) mean ×H L(K1+K2+…+K(r-1)+1) / H Kr +V(C b ) L(K1+K2+…+K(r-1)+2) mean ×H L(K1+K2+…+K(r-1)+2) / H Kr +…+V(C b ) L(K1+K2+…+Kr) mean ×H L(K1+K2+…+Kr) / H Kr )×K r / (K1+K2+…+K r ); Among them, V(C b ) L(K1+K2+…+K(r-1)+2) mean Indicates L K1+K2+…+K(r-1)+Kr Category corresponds to the product characteristics of all deployment points to which the time series model belongs C b Average value; H L(K1+K2+…+Kr) Indicates L K1+K2+…+K(r-1)+Kr The categories correspond to the number of all time series models; H KrK r The categories correspond to the number of all time series models; Calculate representative values ​​for all product features at different deployment points.

[0031] A preliminary classification is performed based on the order p of the autoregressive term and the order q of the moving average term. Different p and q values ​​determine the structure and characteristics of the time series model. Models with the same p and q have similar dynamic behaviors and prediction capabilities to a certain extent. By classifying them into one category, the models can be preliminarily grouped for subsequent more refined classification. Further classification can analyze the distribution and change patterns of product characteristics at different deployment points, and also provide a more accurate classification basis for the subsequent calculation of representative values.

[0032] The representative value of each product characteristic is calculated by weighted summation. First, for each category, the average product characteristic value of the deployment point corresponding to the category is multiplied by the ratio of the number of models in the category to the total number of models in the category, and then multiplied by the weight corresponding to the category. Finally, the weighted results of all categories are added together to obtain the representative value of the product characteristic. Comprehensively considering the contribution of different category models to the representative value, the representative value can more accurately reflect the overall characteristics of the fly ash resource product.

[0033] In this example, in a certain city, the environmental protection department hopes to monitor the characteristics of fly ash resource products to evaluate their environmental impact and resource utilization effect. A fly ash treatment plant was selected and multiple IoT sensors were deployed for data collection; Step 1: According to the characteristics of the fly ash resource product to be detected [heavy metal content, moisture content, particle size], select the corresponding IoT sensor for deployment; select multiple IoT sensor deployment points: [fly ash storage area, fly ash processing equipment outlet, fly ash product storage area, ventilation port, entrance location, center location]; Set the sensor data collection time interval to 1h; build a database to collect all data collected during the 0-4h period; for the fly ash storage area at the deployment point, the corresponding data is expressed as: [{0h: V (heavy metal content), V (moisture content), V (particle size)}, {1h: V (heavy metal content), V (moisture content), V (particle size), …, {4h: V (heavy metal content), V (moisture content), V (particle size)}]; collect the data collected at all deployment points; Step 2, identifying and processing missing values ​​and outliers in the database; For the data on product characteristic moisture content collected at the fly ash storage area at the deployment point, the missing values ​​were preliminarily processed using linear interpolation; Process missing values ​​in the data collected at each deployment point; Set the threshold K and use the Z-score method to identify outliers. For outliers, use the average of neighboring values ​​to replace the outliers for preliminary processing. Identify and process outliers in the data collected at each deployment point; For the product characteristic moisture content collected in the fly ash storage area at the deployment point, an ARMA time series model is constructed: the input is [V (moisture content) 0h ,V(water content) 1h ,…,V(water content) 4h ]; Use ACF and PACF plots to determine AR order p and MA order q; Establish ARMA model formula; Fit ARMA model according to input and estimate parameters by maximum likelihood estimation method; In the trained model, use the model to predict the value of the missing position to interpolate the missing values; For the identified outliers, use the model to predict their normal values ​​for replacement; Different ARMA time series model formulas are obtained by constructing time series models for different product characteristics at different deployment points.

[0034] Step 3: For the characteristic moisture content of products at different deployment points, there are three different ARMA time series model formulas; the six formulas are preliminarily classified according to the order p of the autoregressive term and the order q of the moving average term: the formulas with the same p and the same q are classified into one category; For the formulas after preliminary classification, reclassify them: for the formulas with the same p and q, construct vectors for the ARMA time series model formula parameters; set the number of clusters to 2, and use K-means clustering to divide the ARMA time series model formula parameter vectors corresponding to the moisture content of product characteristics at different deployment points in the same preliminary classification into 2 classes; According to the different results of the preliminary classification, the formula parameters were divided into 2 categories: [L1, L2] and 2 categories: [L3, L4]; the six ARMA time series models whose formula parameters corresponded to the moisture content of product characteristics at different deployment points were also divided into corresponding categories; Category L1: Moisture content time series model in fly ash storage area; Category L2: Moisture content time series model at ventilation outlet; Category L3: Moisture content time series model at fly ash handling equipment outlet and moisture content time series model at inlet; Category L4: Moisture content time series model in fly ash product storage area and moisture content time series model at center location; At 4 h, the moisture content values ​​at each deployment point were: [fly ash storage area: 8%, ventilation outlet: 6%, fly ash handling equipment outlet: 6.6%, inlet location: 7.2%, fly ash product storage area: 5.8%, center location: 5.4%]; The ARMA time series models corresponding to different product characteristics at different deployment points are classified.

[0035] Step S4, for the characteristic moisture content of the product at different deployment points at 4h, calculate the representative value of the characteristic: R = (8% × 1 / 2 + 6% × 1 / 2) × 2 / 6 + (6.9% × 2 / 4 + 5.6% × 2 / 4) × 4 / 6 = 6.5%; Calculate representative values ​​for all product features at different deployment points.

[0036] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An intelligent detection system for fly ash resource products, characterized in that: The system includes: a data acquisition module, a data processing module, a model classification module and a representative value calculation module; The data acquisition module is used to deploy sensors to collect fly ash resource product characteristic data and store it in the database; the data processing module is used to identify and process missing values ​​and abnormal values ​​in the data, and use the time series model to improve the data quality; the model classification module is used to classify the time series models of different deployment points; the representative value calculation module is used to calculate the representative values ​​of product characteristics at different deployment points; The output end of the data acquisition module is connected to the input end of the data processing module; the output end of the data processing module is connected to the input end of the model classification module; the output end of the model classification module is connected to the input end of the representative value calculation module.

2. The intelligent detection system for fly ash resource products according to claim 1 is characterized by: The data acquisition module includes: a sensor deployment unit, a data acquisition unit, and a database unit; The sensor deployment unit is used to deploy IoT sensors at multiple locations; the data acquisition unit is used to collect fly ash resource product characteristic data at set time intervals; the database unit is used to build a database to store data and provide raw data for subsequent analysis; The output end of the sensor deployment unit is connected to the input end of the data acquisition unit; the output end of the data acquisition unit is connected to the input end of the database unit; the output end of the database unit is connected to the input end of the data processing module.

3. The intelligent detection system for fly ash resource products according to claim 2 is characterized in that: The data processing module includes: a preliminary processing unit, a time series model unit and a data processing unit; The preliminary processing unit is used to identify and preliminarily process missing values ​​and abnormal values ​​in the data; the time series model unit is used to construct a time series model based on the preliminarily processed data; the data processing unit is used to further interpolate and replace data based on the time series model; The output end of the preliminary processing unit is connected to the input end of the timing model unit; the output end of the timing model unit is connected to the input end of the data processing unit; the output end of the data processing unit is connected to the input end of the model classification module.

4. The intelligent detection system for fly ash resource products according to claim 3 is characterized by: The model classification module includes: a preliminary classification unit, a cluster number setting unit and a model classification unit; The preliminary classification unit is used to preliminarily classify the model according to the order of the autoregressive term and the moving average term; the cluster number setting unit is used to set the cluster number of the K-means clustering model; the model classification unit is used to classify the model parameter vector according to the K-means clustering, and classify the models of different deployment points into different categories; The output end of the preliminary classification unit is connected to the input end of the cluster number setting unit; the output end of the cluster number setting unit is connected to the input end of the model classification unit; the output end of the model classification unit is connected to the input end of the representative value calculation unit.

5. The intelligent detection system for fly ash resource products according to claim 4 is characterized in that: The representative value calculation module includes: a representative value calculation unit; The representative value calculation unit is used to calculate the representative values ​​of product characteristics of different deployment points according to the model classification results, and perform weighted calculation by comprehensively considering the characteristic average values ​​of the deployment points to which models of different categories belong and the number of models.

6. An intelligent detection method for fly ash resource products, characterized in that: The method comprises the following steps: Step 1: Deploy IoT sensors, set the collection time interval, and build a database to collect fly ash resource product characteristic data at different times at each deployment point; Step 2: Linearly interpolate missing values, identify and process outliers, and build an ARMA model to interpolate missing values ​​and replace outliers; Step 3: Preliminarily classify the ARMA models by order, and then use K-means clustering to classify the model parameter vectors, so as to classify the models into corresponding categories; Step 4: Calculate the representative values ​​of product characteristics at different deployment points based on the average characteristics of the deployment points to which models of different categories belong and the number of models.

7. An intelligent detection method for fly ash resource products according to claim 6, characterized in that: In step 1, according to the characteristics of the fly ash resource product to be tested [C1, C2, …, C u ], select the corresponding IoT sensor for deployment; select multiple IoT sensor deployment points: [D1,D2,…,D v ]; Where u is a positive integer, indicating the number of fly ash resource product characteristics; C u represents the characteristics of the u-th fly ash resource product; v is a positive integer, indicating the number of IoT sensor deployment points; D v represents the vth IoT sensor deployment point; Set the sensor data collection time interval to T; build a database to collect all the data collected during the period 0 to w·T; for the deployment point Da, the corresponding data is expressed as: [{t0:V(C1),V(C2),…,V(C u )},{t1:V(C1),V(C2),…,V(C u )},…,{t w :V(C1),V(C2),…,V(C u )}]; collect data collected from all deployment points; Where w is a positive integer, indicating the number of times data is collected; a is a positive integer, a∈{1,2,…,v}, D a represents the ath IoT sensor deployment point; t w represents the time w·T; V(C u ) represents the value of the characteristic of the u-th fly ash resource product.

8. An intelligent detection method for fly ash resource products according to claim 7, characterized in that: In step 2, missing values ​​and outliers in the database are identified and processed; For deployment point D a Collected product characteristics C b Data, t i The missing values ​​at the moment are preliminarily processed using linear interpolation: V(C b ) ti insert =V(C b ) t(i-1) +(V(C b ) t(i+1) -V(C b ) t(i-1) ) / (t (i+1) -t (i-1) )·(t i -t (i-1) ); Where i is a positive integer, representing the time sequence, i∈{1,2,…,w-1}; b is a positive integer, b∈{1,2,…,u}, C b represents the characteristics of the bth fly ash resource product; V(C b ) ti insert Indicates t i The interpolation value of the fly ash resource product characteristics at time b; V(C b ) t(i+1) Indicates t (i+1) The collected values ​​of the characteristics of the fly ash resource product at time b; Process missing values ​​in the data collected at each deployment point; Use the Z-score method to identify outliers; for deployment point D a Collected product characteristics C b , t j The data V(C b ) tj :Z(C b ) tj =(V(C b ) tj -μ(C b )) / σ(C b ); Set the threshold K, when |Z(C b ) tj When |>K, the value is considered an outlier; For outliers, use the average value of neighboring values ​​to replace the outliers for preliminary processing: V(C b ) tj instead =(V(C b ) t(j-1) +V(C b ) t(j+1) ) / 2; Where j is a positive integer, representing a time sequence, j∈{1,2,…,w-1}; V(C b ) tj instead Indicates t j The replacement value of the fly ash resource product characteristics at time b; V(C b ) t(j+1) Indicates t (j+1) The collection value of the characteristics of the fly ash resource product at the bth time; μ(C b ) is product characteristic C b The mean value of σ(C b ) is product characteristic C b The standard deviation of Identify and process outliers in the data collected at each deployment point; For deployment point D a Collected product characteristics C b , construct an ARMA time series model: Input is [V(C b ) t0 ,V(C b ) t1 ,…,V(C b ) tw ]; Use ACF and PACF plots to determine AR order p and MA order q; Establish an ARMA model: V(C b ) ts =ϕ1V(C b ) t(s−1) +ϕ2V(C b ) t(s−2) +…+ϕ p V(C b ) t(s−p) +θ1ε t(s−1) +θ2ε t(s−2) +…+θ q ε t(s−q) +ε ts ; Fit the ARMA model according to the input and estimate the parameters by the maximum likelihood estimation method; In the trained model, use the model to predict the value of the missing position to interpolate the missing values; For the identified outliers, use the model to predict their normal values ​​for replacement; Where s is a positive integer, representing the time series, s ≥ p, s ≥ q; p represents the order of the autoregressive term; q represents the order of the moving average term; ϕ1, ϕ2, …, ϕ p Represents the parameters of the autoregressive part AR; θ1, θ2, …, θ q Represents the parameters of the moving average MA part; ε ts represents random error; Different ARMA time series model formulas are obtained by constructing time series models for different product characteristics at different deployment points.

9. The intelligent detection method for fly ash resource products according to claim 8, characterized in that: In step 3, for different deployment point product features C b , there are v different ARMA time series model formulas; The v formulas are preliminarily classified according to the order p of the autoregressive term and the order q of the moving average term: formulas with the same p and formulas with the same q are classified into one category; For the formulas after preliminary classification, reclassify them: for formulas with the same p and q, the parameters ϕ1, ϕ2, …, ϕ p and θ1,θ2,…,θ q Construct the vector [ϕ1,ϕ2,…,ϕ p ,θ1,θ2,…,θ q ]; Set the number of clusters K1, and use K-means clustering to cluster the same preliminary classification of different deployment point product features C b The corresponding ARMA time series model formula parameter vector [ϕ1,ϕ2,…,ϕ p ,θ1,θ2,…,θ q ] is divided into K1 classes; According to the different results of the preliminary classification, the formula parameters are divided into K1 categories: [L1, L2, ..., L K1 ], K2 classes: [L K1+1 ,L K1+2 ,…,L K1+K2 ], ..., K r Class: [L K1+K2+…+K(r-1)+1 ,L K1+K2+…+K(r-1)+2 ,…,L K1+K2+…+K(r-1)+Kr ]; formula parameters correspond to product features at different deployment points C b The v ARMA time series models are also divided into corresponding categories; Among them, r is a positive integer, which indicates the number of types of preliminary classification; L K1+K2+…+K(r-1)+Kr Indicates K1+K2+…+K r categories; The ARMA time series models corresponding to different product characteristics at different deployment points are classified.

10. An intelligent detection method for fly ash resource products according to claim 9, characterized in that: In step S4, for the product characteristics C at different deployment points at a certain moment b , calculate the representative value of this characteristic: R=(V(C b ) L1 mean ×H L1 / H K1 +V(C b ) L2 mean ×H L2 / H K1 +…+V(C b ) LK1 mean ×H LK1 / H K1 )×K1 / (K1+K2+…+K r )+(V(C b ) L(K1+1) mean ×H L(K1+1) / H K2 +V(C b ) L(K1+2) mean ×H L(K1+2) / H K2 +…+V(C b ) L(K1+K2) mean ×H L(K1+K2) / H K2 )×K2 / (K1+K2+…+K r )+…+(V(C b ) L(K1+K2+…+K(r-1)+1) mean ×H L(K1+K2+…+K(r-1)+1) / H Kr +V(C b ) L(K1+K2+…+K(r-1)+2) mean ×H L(K1+K2+…+K(r-1)+2) / H Kr +…+V(C b ) L(K1+K2+…+Kr) mean ×H L(K1+K2+…+Kr) / H Kr )×K r / (K1+K2+…+K r ); Among them, V(C b ) L(K1+K2+…+K(r-1)+2) mean Indicates L K1+K2+…+K(r-1)+Kr Category corresponds to the product characteristics of all deployment points to which the time series model belongs C b Average value; H L(K1+K2+…+Kr) Indicates L K1+K2+…+K(r-1)+Kr The categories correspond to the number of all time series models; H Kr K r The categories correspond to the number of all time series models; Calculate representative values ​​for all product features at different deployment points.

Images