Search

CN-122020242-A - Intelligent hazardous waste classification processing method based on multi-parameter fusion

CN122020242ACN 122020242 ACN122020242 ACN 122020242ACN-122020242-A

Abstract

The invention relates to the technical field of hazardous waste treatment and discloses an intelligent hazardous waste classification treatment method based on multi-parameter fusion, which comprises the following steps of collecting current multi-modal data of hazardous waste and loading historical multi-modal data; the multi-mode data comprises chemical component data, 400-2500nm wave band images, molecular vibration spectrums and combustion heat values, wherein a classification model is called to perform cross-mode feature fusion by combining current multi-mode data and historical multi-mode data, a maximum correlation coefficient feature pair is extracted, a linear programming model is established based on the maximum correlation coefficient feature pair, the linear programming model is used for calculating the predicted emission amount of the dioxin and optimizing the recovery rate of valuable metals, the actual emission amount of the dioxin is monitored in real time, and if the deviation between the actual emission amount of the dioxin and the predicted emission amount of the dioxin is larger than a specified threshold value, the classification calibration is triggered. According to the technical scheme, misjudgment caused by single information in the traditional method can be effectively avoided, and the recognition accuracy is improved.

Inventors

  • ZHANG JINGZHENG
  • XU PENG
  • HONG YICHANG
  • MA JUN
  • TAN WENCHAO
  • ZHANG XIAOQING
  • TAO WEIZHONG

Assignees

  • 中国电建集团贵阳勘测设计研究院有限公司

Dates

Publication Date
20260512
Application Date
20251231

Claims (8)

  1. 1. The intelligent classification treatment method for the hazardous waste based on multi-parameter fusion is characterized by comprising the following steps of: collecting current multi-mode data of dangerous waste, and loading historical multi-mode data, wherein the multi-mode data comprises chemical component data, 400-2500nm wave band images, molecular vibration spectrums and combustion heat values; Combining the current multi-modal data and the historical multi-modal data, calling a classification model to perform cross-modal feature fusion, and extracting a maximum correlation coefficient feature pair; Establishing a linear programming model based on the maximum correlation coefficient characteristic pair, wherein the linear programming model is used for calculating the predicted emission quantity of dioxin and optimizing the recovery rate of valuable metals; and monitoring the actual emission amount of the dioxin in real time, and triggering classification calibration if the deviation between the actual emission amount of the dioxin and the predicted emission amount of the dioxin is larger than a specified threshold value.
  2. 2. The intelligent classification processing method for the hazardous waste based on multi-parameter fusion is characterized in that the chemical component data are obtained through GC-MS, the 400-2500nm wave band image is obtained through a hyperspectral camera, the molecular vibration spectrum is obtained through a near infrared spectrometer, and the combustion heat value is obtained through a heat value sensor.
  3. 3. The intelligent classification processing method for hazardous waste based on multi-parameter fusion according to claim 1, wherein performing the cross-modal feature fusion comprises the steps of: Judging whether concept drift occurs, and if so, updating a classification model by adopting an incremental random forest model; Carrying out typical correlation analysis on different types of data, and extracting a maximum correlation coefficient characteristic pair; And reducing the dimension of the maximum correlation coefficient characteristic pair input sparse self-encoder to 50 dimensions.
  4. 4. The intelligent classification and processing method for hazardous waste based on multi-parameter fusion according to claim 3, wherein the objective function of the canonical correlation analysis is that , The constraint is that Wherein, the method comprises the steps of, Is the output control quantity of a PID controller, Is a linear combination coefficient, Σxy is a covariance matrix of chemical components and spectral data.
  5. 5. The intelligent classification processing method for dangerous waste based on multi-parameter fusion according to claim 1, wherein the objective function of the linear programming model is Max (0.6×heat value stability+0.3×valuable metal recovery rate-0.1×dioxin predicted emission); The constraint condition of the linear programming model is that the heat value fluctuation is less than or equal to +/-5%, cl is less than or equal to 1.5wt% and S is less than or equal to 0.8wt% and accords with the incompatible waste isolation rule.
  6. 6. The intelligent classification processing method for hazardous waste based on multi-parameter fusion according to claim 5, wherein the calculation of the predicted emission amount of dioxin is realized through a time sequence convolution network, and the calculated input parameters are the incineration temperature T, the chlorine content cl% and the catalyst metal content cu%, expressed as: ; Optimizing the recovery rate of valuable metals is realized by gradient lifting decision tree, and input characteristics comprise initial concentration and pH value of Au, ag and Pd in waste.
  7. 7. The intelligent classification and processing method for dangerous waste based on multi-parameter fusion according to claim 1, wherein the classification and calibration means that the control of the incinerator is adjusted through a PID controller to realize the adjustment of the dioxin emission and the analysis is restarted from the cross-modal feature fusion step; The PID controller adjusts the control of the incinerator, expressed as: ; Where e (t) is the actual and predicted emissions of dioxin, and Kp, ki and Kd are control coefficients, respectively.
  8. 8. The intelligent classification processing method for dangerous waste based on multi-parameter fusion according to claim 2, wherein the 400-2500nm wave band images are subjected to frequency domain decomposition, low frequency components represent matrix uniformity, and high frequency components represent images of microscale pollutant aggregation in the same wave band, so that spectral characteristics of the waste can be reflected.

Description

Intelligent hazardous waste classification processing method based on multi-parameter fusion Technical Field The invention relates to the technical field of hazardous waste treatment, in particular to an intelligent hazardous waste classification treatment method based on multi-parameter fusion. Background At present, the production of dangerous wastes in China is in an increasing situation year by year, but the management system is not perfect, blind areas exist in the supervision coverage, and illegal dumping and illegal disposal events still occur. The hazardous waste classification treatment, the traditional technical means face a plurality of challenges to be solved urgently. In order to suppress mess, the regulatory department has explicitly proposed the urgent requirement of establishing the traceable, accurate management system of the whole process, need to realize from producing, transporting the high-efficient management and control of the processing link. At present, manual experience judgment and a single detection technology are generally relied on in the industry, for example, patent CN115722460B only adopts XRF to analyze heavy metal components, serious data island problems exist in the mode, complex characteristics of dangerous wastes are difficult to comprehensively capture, according to NISTSRM8692 standard sample test verification, the mixed waste misjudgment rate is higher due to a single detection means, the situation that copper-containing sludge is misjudged as organic waste liquid occurs in actual application, the fluctuation range of heat value after compatibility is large, and the stability and safety of a treatment process are seriously affected. Disclosure of Invention In order to solve the problems, the application provides an intelligent classification treatment method for dangerous wastes based on multi-parameter fusion, which comprises the following steps: collecting current multi-mode data of dangerous waste, and loading historical multi-mode data, wherein the multi-mode data comprises chemical component data, 400-2500nm wave band images, molecular vibration spectrums and combustion heat values; Combining the current multi-modal data and the historical multi-modal data, calling a classification model to perform cross-modal feature fusion, and extracting a maximum correlation coefficient feature pair; based on the maximum correlation coefficient characteristic pair, a linear programming model is established, and the linear programming model is used for calculating the predicted emission quantity of dioxin and optimizing the recovery rate of valuable metals; And monitoring the actual emission amount of the dioxin in real time, and triggering classification calibration if the deviation between the actual emission amount of the dioxin and the predicted emission amount of the dioxin is larger than a specified threshold value. The chemical composition data are obtained through GC-MS, the 400-2500nm wave band images are obtained through a hyperspectral camera, the molecular vibration spectrum is obtained through a near infrared spectrometer, and the combustion heat value is obtained through a heat value sensor. And carrying out frequency domain decomposition on the 400-2500nm wave band image, wherein the low-frequency component represents matrix uniformity, and the high-frequency component represents the image of microscale pollutant aggregation with the same wave band, so that the spectral characteristics of waste can be reflected. Further, performing the cross-modal feature fusion includes the steps of: Judging whether concept drift occurs, and if so, updating a classification model by adopting an incremental random forest model; Carrying out typical correlation analysis on different types of data, and extracting a maximum correlation coefficient characteristic pair; And reducing the dimension of the maximum correlation coefficient characteristic pair input sparse self-encoder to 50 dimensions. The objective function of a typical correlation analysis is,The constraint is thatWherein, the method comprises the steps of,Is the output control quantity of a PID controller,Is a linear combination coefficient, Σxy is a covariance matrix of chemical components and spectral data. Further, the objective function of the linear programming model is Max (0.6×heating value stability+0.3×valuable metal recovery-0.1×dioxin predicted emission); The constraint condition of the linear programming model is that the heat value fluctuation is less than or equal to +/-5%, cl is less than or equal to 1.5wt% and S is less than or equal to 0.8wt% and accords with the incompatible waste isolation rule. The method comprises the steps of calculating the predicted emission of dioxin through a time sequence convolution network, wherein input parameters are the incineration temperature T, the chlorine content Cl and the catalyst metal content Cu during calculation, and the parameters are expressed as follows: ;