CN-122024904-A - Data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method

CN122024904ACN 122024904 ACN122024904 ACN 122024904ACN-122024904-A

Abstract

The invention discloses a data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method, which belongs to the field of neural networks and comprises the steps of obtaining a microscopic image set of a copper ferrite denitration catalyst before microwave treatment, determining local field intensity distribution feature vectors based on microwave field parameters, extracting initial structure feature vectors from the microscopic image set, matching in a pre-built microwave structure evolution mapping library according to the microwave field parameters to obtain corresponding initial structure evolution trend vectors, adjusting the initial structure evolution trend vectors by taking the local field intensity distribution feature vectors as conditional weights to obtain target structure evolution trend vectors, fusing the target structure evolution trend vectors with the initial structure feature vectors to obtain process structure feature tensors, and inputting the process structure feature tensors into a neural network prediction model to obtain a predicted performance result. The method can improve the accuracy of the performance prediction of the microwave-assisted copper ferrite denitration catalyst.

Inventors

WANG YUANQING
JIA BAOZHU
JIA XIAOPING
JI RAN
ZHAO XU

Assignees

广东海洋大学

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. The method for predicting the performance of the microwave-assisted copper ferrite denitration catalyst based on data driving is characterized by comprising the following steps of: Acquiring a microscopic image set of a copper ferrite denitration catalyst before microwave treatment; Determining a local field intensity distribution feature vector corresponding to a microstructure space of the copper ferrite denitration catalyst based on preset microwave field parameters, extracting an initial structural feature vector of the copper ferrite denitration catalyst from the microscopic image set, matching in a pre-constructed microwave structure evolution mapping library according to the microwave field parameters to obtain a corresponding initial structure evolution trend vector, adjusting the initial structure evolution trend vector by taking the local field intensity distribution feature vector as a conditional weight to obtain a target structure evolution trend vector, and fusing the target structure evolution trend vector with the initial structural feature vector to obtain a process structure feature tensor reflecting interaction of a microwave field and a catalyst structure, wherein the microwave structure evolution mapping library is obtained by processing time sequence structure image sequences under different microwave field parameters; inputting the process structural feature tensor into a pre-trained performance prediction model to obtain a predicted performance result of the copper ferrite denitration catalyst after the treatment is completed under the microwave field parameters.
2. The method for predicting performance of a data-driven microwave-assisted copper ferrite denitration catalyst according to claim 1, wherein the performance prediction model comprises a space-time feature encoder, a multi-head attention fusion layer and a regression prediction layer which are sequentially connected, the process structure feature tensor is input into a pre-trained performance prediction model, and a predicted performance result of the copper ferrite denitration catalyst after the copper ferrite denitration catalyst is processed under the microwave field parameters is obtained, and the method comprises the following steps: Inputting the process structure feature tensor into a pre-trained performance prediction model, and extracting local structure features and space-time dependency relations in the process structure feature tensor through the space-time feature encoder to construct a coding result; processing the coding result through the multi-head attention fusion layer to obtain a dynamic evolution trend vector; And carrying out nonlinear mapping on the dynamic evolution trend vector through the regression prediction layer, and outputting the prediction performance result, wherein the prediction performance result comprises denitration efficiency and catalyst stability indexes.
3. The method for predicting performance of a data-driven microwave-assisted copper ferrite denitration catalyst according to claim 1, wherein determining a local field intensity distribution eigenvector corresponding to a microstructure space of the copper ferrite denitration catalyst based on preset microwave field parameters comprises: Normalizing the microwave field parameters to obtain normalized microwave parameters, wherein the microwave field parameters comprise microwave frequency, microwave power and microwave irradiation time; calculating a reference field intensity vector reflecting the average field intensity level based on the normalized microwave parameters; according to the porosity and component distribution information reflected by the microscopic image set, estimating and obtaining dielectric constant distribution feature vectors of each position of the microstructure space through finite element electromagnetic simulation; And performing element-wise multiplication operation on the reference field intensity vector and the dielectric constant distribution characteristic vector to generate the local field intensity distribution characteristic vector.
4. The method for predicting performance of a data-driven microwave-assisted copper ferrite denitration catalyst according to claim 3, wherein the performing element-wise multiplication operation on the reference field intensity vector and the dielectric constant distribution feature vector to generate the local field intensity distribution feature vector includes: calculating the space autocorrelation coefficient of the dielectric constant distribution characteristic vector; Performing weighted smoothing processing on the dielectric constant distribution feature vector based on the spatial autocorrelation coefficient to generate an associated feature vector reflecting spatial dependence; And multiplying the reference field intensity vector and the associated feature vector by elements to generate the local field intensity distribution feature vector.
5. The method for predicting performance of a data-driven microwave-assisted copper ferrite denitration catalyst as claimed in claim 1, wherein the extracting the initial structural feature vector of the copper ferrite denitration catalyst from the microscopic image set comprises: Preprocessing the microscopic image set to obtain a standardized microscopic image set, wherein the microscopic image set comprises a scanning electron microscope image and a transmission electron microscope image; And extracting multidimensional morphological characteristics from the standardized microscopic image set based on an image processing algorithm to form an initial structural characteristic vector of the copper ferrite denitration catalyst, wherein the morphological characteristics comprise average particle size, size distribution variance, porosity and shape factor.
6. The data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method according to claim 1, wherein the matching is performed in a pre-constructed microwave structure evolution mapping library according to the microwave field parameters to obtain a corresponding initial structure evolution trend vector, and the method comprises the following steps: Matching the microwave field parameters with index keys prestored in the microwave structure evolution mapping library to obtain a matching result; Determining an associated sequence of time-sequential structural images based on the matching result; calculating a differential vector sequence of the structure characteristics between adjacent time frames from the time sequence structure image sequence; and performing time sequence fitting on the differential vector sequence, extracting characteristic parameters corresponding to the variation trend, and forming the initial structure evolution trend vector.
7. The data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method as claimed in claim 6, wherein the microwave structure evolution mapping library is obtained through training of time sequence structure image sequences under different microwave field parameters, and comprises the following steps: respectively carrying out microwave treatment on an initial copper ferrite denitration catalyst sample by utilizing a plurality of initial microwave field parameter combinations, and collecting an initial time sequence structure image sequence which changes with time in the treatment process; image denoising and feature alignment are carried out on the initial time sequence structure image sequence, and a processing result is obtained; combining the initial microwave field parameters as an index key, and establishing a one-to-one mapping relation by taking the corresponding processing result as a data value; And generating the microwave structure evolution mapping library taking the index key as a retrieval basis based on all the established mapping relations.
8. The method for predicting performance of a data-driven microwave-assisted copper ferrite denitration catalyst according to claim 1, wherein the adjusting the initial structure evolution trend vector with the local field intensity distribution feature vector as a conditional weight to obtain a target structure evolution trend vector comprises: Calculating weight distribution of the initial structure evolution trend vector based on the local field intensity distribution feature vector; And carrying out weighted fusion on the initial structure evolution trend vector according to the weight distribution to obtain the target structure evolution trend vector.
9. The method for predicting performance of a data-driven microwave-assisted copper ferrite denitration catalyst according to claim 1, wherein the fusing the target structure evolution trend vector with the initial structure feature vector to obtain a process structure feature tensor comprises: Performing tensor splicing operation on the target structure evolution trend vector and the initial structure feature vector to obtain a spliced feature tensor; and carrying out feature fusion on the spliced feature tensor, and outputting the process structure feature tensor.
10. The data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method according to claim 1, wherein the training process of the performance prediction model comprises the following steps: obtaining a training set, wherein the training set comprises a plurality of samples, and each sample comprises an initial process structure feature tensor and a corresponding real performance label; Inputting the training set into an initial performance prediction model, optimizing parameters of the initial performance prediction model through a gradient descent algorithm by adopting a multi-task joint learning strategy so as to minimize an overall loss function, and determining the performance prediction model based on the optimized model parameters.

Description

Data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method Technical Field The invention relates to the field of neural networks, in particular to a data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method. Background The microwave auxiliary treatment can obviously regulate and control the microstructure (such as particle size, pores and defect state) of the copper ferrite denitration catalyst by means of selective heating, changing a local electric field and the like, so that the denitration activity and stability of the copper ferrite denitration catalyst are optimized. However, the effect of the microwaves and the catalyst is a dynamic coupling process involving electromagnetic fields, thermal effects and material responses, and the conventional method is difficult to accurately describe the internal law, so that the method for measuring the final performance of the microwave treatment is difficult to get rid of an inefficient 'trial-and-error' research and development mode, and further, the directional design and process optimization of the performance of the catalyst are difficult to realize. The prior art generally employs a static correlation method based on results by first fixing a set of microwave parameters (e.g., power, frequency, time) to treat the catalyst, then characterizing only the final microstructure after treatment (e.g., scanning electron microscope image), and relying on the characterization results to perform performance prediction. The fundamental disadvantage of this approach is that it completely ignores the dynamic evolution path of the microstructure during microwave processing, as well as the spatial non-uniformity of the interaction of the electromagnetic field with the material, and thus does not reliably guide the prediction of the performance of new parameters or new material systems. Disclosure of Invention The invention provides a data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method, which can improve the accuracy of microwave-assisted copper ferrite denitration catalyst performance prediction. The embodiment of the invention provides a data-driven microwave-assisted copper ferrite denitration catalyst performance prediction method, which comprises the following steps: Acquiring a microscopic image set of a copper ferrite denitration catalyst before microwave treatment; Determining a local field intensity distribution feature vector corresponding to a microstructure space of the copper ferrite denitration catalyst based on preset microwave field parameters, extracting an initial structural feature vector of the copper ferrite denitration catalyst from the microscopic image set, matching in a pre-constructed microwave structure evolution mapping library according to the microwave field parameters to obtain a corresponding initial structure evolution trend vector, adjusting the initial structure evolution trend vector by taking the local field intensity distribution feature vector as a conditional weight to obtain a target structure evolution trend vector, and fusing the target structure evolution trend vector with the initial structural feature vector to obtain a process structure feature tensor reflecting interaction of a microwave field and a catalyst structure, wherein the microwave structure evolution mapping library is obtained by processing time sequence structure image sequences under different microwave field parameters; inputting the process structural feature tensor into a pre-trained performance prediction model to obtain a predicted performance result of the copper ferrite denitration catalyst after the treatment is completed under the microwave field parameters. According to the embodiment of the invention, the accurate material starting point for prediction is established by acquiring the microscopic image set before microwave processing and extracting the initial structural feature vector, so that the prediction deviation caused by neglecting the initial structural difference is avoided. The method accurately characterizes the non-uniform physical nature of the interaction of microwaves and catalysts based on the microwave field parameters to determine the local field intensity distribution feature vector corresponding to the microstructure space, converts macroscopic microwave parameters into real local driving conditions affecting the structural evolution, and fundamentally overcomes model errors caused by regarding the microwave field as a uniform field. According to the method, an initial structure evolution trend vector is obtained by matching microwave field parameters from a pre-built microwave structure evolution mapping library, and the method introduces a universal dynamic rule of structure evolution under specific microwave parameters revealed by historical experimental data, so that prediction is not limited to a single final