# US 2020/0394494 A1 · Method And System For Predicting Content Of Multiple Components In Rare Earth Extraction Process

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2020/0394494 A1
- Document ID: US20200394494A1
- Status: Published application
- Kind: A1
- Application number: 16/564,430
- Filed: 2019-09-09
- Published: 2020-12-17
- Earliest claimed: 2019-06-17
- Assignees: East China Jiaotong University
- Applicants: East China Jiaotong University
- Inventors: Rongxiu Lu; Quanheng He; Hui Yang; Jianyong Zhu; Gang Yang; Fangping Xu
- Web page: https://patentgrep.com/patent/US20200394494A1

## CPC

- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 3/0499: Feedforward networks
- G06N 3/086: using evolutionary algorithms, e.g. genetic algorithms or genetic programming
- G06N 3/09: Supervised learning
- G06N 3/0985: Hyperparameter optimisation; Meta-learning; Learning-to-learn
- C22B 59/00: Obtaining rare earth metals
- Y02P 10/20: Recycling
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- C22B: PRODUCTION AND REFINING OF METALS (electrolytic C25); PRETREATMENT OF RAW MATERIALS
- Y02P: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS

## Abstract

Described is a method for predicting multiple components' content in a case that rare earth ions with and without color feature coexist, and relates to component content prediction in rare earth extraction process. It is difficult to quickly accurately detect component's content in rare earth extraction process. Because of relatively large difference between images' color features of CePr/Nd mixed solution with colorless Ce ions and Pr/Nd solution, detecting content method of single rare earth element based on color feature is no longer applicable. The method includes: first searching for H and S components with maximum correlation with component content in HSI color space; establishing ELM based multi-component content soft measurement model using H and S component first-order moment as input; and for uncertainty of initial weight and ELM (extreme learning machine) model's threshold, optimizing model parameters using genetic algorithm GA to optimize ELM model for component content prediction higher precision.

## Claims

### Claim 1

A method for predicting content of multiple components in a rare earth extraction process, wherein the prediction method comprises:
obtaining color characteristic values, in different color space, of an image of a cerium praseodymium/neodymium mixed solution in which rare earth ions with a color feature and rare earth ions without a color feature coexist;
determining, in the different color space according to the color characteristic values of the image of the cerium praseodymium/neodymium mixed solution and known content of cerium praseodymium/neodymium components, HSI color feature components with a maximum correlation;
obtaining an H and S component first-order moment with a maximum correlation in the HSI color feature components;
establishing an extreme learning machine based multi-component content soft measurement model by using the H and S component first-order moment as input and by using the known content values of the cerium praseodymium/neodymium components as output;
optimizing an initial weight matrix and a hidden-layer threshold matrix of the extreme learning machine based multi-component content soft measurement model to obtain a genetic algorithm-extreme learning machine based multi-component content soft measurement model; and
determining, according to the genetic algorithm-extreme learning machine based multi-component content soft measurement model, content values of multiple elements in a to-be-detected rare earth mixed solution by using the H and S component first-order moment as input.

### Claim 2 (depends on 1)

The method for predicting content of multiple components in a rare earth extraction process according to claim 1, wherein the extracting, in different color space color, color characteristic values of an image of a cerium praseodymium/neodymium mixed solution in which rare earth ions with a color feature and rare earth ions without a color feature coexist specifically comprises:
obtaining, by using a machine vision technology, the image of the cerium praseodymium/neodymium mixed solution in which rare earth ions with a color feature and rare earth ions without a color feature coexist; and
obtaining the color characteristic values of the image of the cerium praseodymium/neodymium mixed solution image in the different color space based on the image of the cerium praseodymium/neodymium mixed solution.

### Claim 3 (depends on 1)

The method for predicting content of multiple components in a rare earth extraction process according to claim 1, wherein the extreme learning machine based multi-component content soft measurement model is specifically represented as follows:
the extreme learning machine based multi-component content soft measurement model comprises an input layer, a hidden layer, and an output layer;
when an input sample is given, a matrix calculation formula of a hidden-layer neuron output is H=g(WXT+b), wherein H is the hidden-layer neuron output; g is an activation function of a hidden-layer neuron; X is the H and S component first-order moment; W is a weight matrix between the output layer and the hidden layer; and b is a hidden-layer neuron threshold matrix; and
a calculation formula of a neural network output is P=(HT β), wherein β is a weight matrix from the hidden layer to the output layer; an ELM model can be uniquely determined provided that β is determined; and when an output sample Y is given, the weight matrix can be solved by using
min
β


H
T

β
-
Y

,
and a solution of the weight matrix is {circumflex over (β)}=(HT)+Y, wherein (HT)30 is a pseudo-inverse matrix of a transposed matrix HT.

### Claim 4 (depends on 3)

The method for predicting content of multiple components in a rare earth extraction process according to claim 3, wherein the optimizing an initial weight matrix and a hidden-layer threshold matrix of the extreme learning machine based multi-component content soft measurement model to obtain a genetic algorithm-extreme learning machine based multi-component content soft measurement model specifically comprises:
using the first-order moments of the H component and the S component and the known content of the cerium praseodymium/neodymium component as training set data and test set data;
conducting normalization processing on the training set data and the test set data;
optimizing, according to the training set data and the test set data having undergone normalization processing, the weight matrix between the output layer and the hidden layer and the hidden-layer threshold matrix by using a genetic algorithm, to obtain an optimal weight between the output layer and the hidden layer and an optimal hidden-layer threshold;
determining both an optimal weight between the output layer and the hidden layer and an optimal hidden-layer neuron threshold by using the genetic algorithm, and assigning the optimal weight and the optimal hidden-layer neuron threshold to an extreme learning machine as an initial weight and a threshold, and setting a number of hidden-layer nodes; and
determining the genetic algorithm-extreme learning machine based multi-component content soft measurement model by using the initial weight, the threshold, and the hidden-layer node.

### Claim 5

A system for predicting content of multiple components in a rare earth extraction process, wherein the system comprises:
a color characteristic value obtaining module, configured to obtain color characteristic values, in different color space, of an image of a cerium praseodymium/neodymium mixed solution in which rare earth ions with a color feature and rare earth ions without a color feature coexist;
an HSI color feature component determining module, configured to determine, in the different color space according to the color characteristic values of the image of the cerium praseodymium/neodymium mixed solution and known content of cerium praseodymium/neodymium components, HSI color feature components with a maximum correlation;
an H and S component first-order moment obtaining module, configured to obtain an H and S component first-order moment with a maximum correlation in the HSI color feature components;
extreme learning machine based multi-component content soft measurement model construction module, configured to establish an extreme learning machine based multi-component content soft measurement model by using the H and S component first-order moment as input and by using the known content values of the cerium praseodymium/neodymium components as output;
a genetic algorithm-extreme learning machine based multi-component content soft measurement model construction module, configured to optimize an initial weight and a hidden-layer threshold of the extreme learning machine based multi-component content soft measurement model to obtain a genetic algorithm-extreme learning machine based multi-component content soft measurement model; and
a multi-component content determining module, configured to determine, according to the genetic algorithm-extreme learning machine based multi-component content soft measurement model, content of multiple components in a to-be-detected rare earth mixed solution by using the H and S component first-order moment as input.

### Claim 6 (depends on 5)

The system for predicting content of multiple components in a rare earth extraction process according to claim 5, the color characteristic value obtaining module specifically comprises:
a solution image obtaining unit, configured to obtain, by using a machine vision technology, the image of the cerium praseodymium/neodymium mixed solution in which rare earth ions with a color feature and rare earth ions without a color feature coexist; and
a color characteristic value determining unit, configured to extract the color characteristic values of the image of the cerium praseodymium/neodymium mixed solution image in the different color space based on the image of the cerium praseodymium/neodymium mixed solution.
