Ore collecting device multi-parameter optimization design method based on deep sea tuberculosis grading

Abstract

The invention discloses an ore collection device multi-parameter optimization design method based on deep sea tuberculosis grading, and the method employs a discrete genetic algorithm to carry out the overall optimization of a plurality of design parameters of an ore collection device, and improves the collection efficiency. An elite retention strategy and a uniform two-point crossover method are introduced, several individuals with the highest fitness are selected as elite individuals, two points are randomly selected from gene strings of two parent individuals according to the uniform two-point crossover method, and then gene segments between the two points are exchanged to generate new individuals, so that exploration of new parameter combinations is facilitated, population diversity is increased, and the method is suitable for large-scale popularization and application. Therefore, a better design solution can be found; according to the method, a maximum likelihood method is adopted to estimate parameters of marginal probability density functions of the size and mass of the nodule, and a size-mass joint probability distribution function of nodule particles is constructed, so that the size and mass distribution of the nodule can be more accurately fitted, and a more real seabed environment model is created for the subsequent design of a mineral collection device; and relatively real seabed tuberculosis particle distribution is provided.

Description

Technical Field

[0001] The invention belongs to the technical field of deep-sea mining, and in particular relates to a multi-parameter optimization design method for a ore collection device based on deep-sea nodule grading. Background Art

[0002] In the field of engineering, especially in the development of deep-sea mining technology, accurate prediction of environmental factors and material properties is crucial. These factors are often random variables, may be correlated with each other, and have a significant impact on engineering design and performance. In the design of deep-sea mining robots, input random variables such as ore particle size or particle density and environmental factors are often correlated. The existence of these correlations is crucial for the accurate solution of the reliability-based design optimization (RBDO) problem. In order to solve this problem, it is necessary to know the joint cumulative distribution function (CDF) of these correlated input random variables. Traditional RBDO technology mainly relies on accurately observed data, but actual environmental or experimental data are often not fully reported due to the limitations of measurement equipment and experimental characteristics, that is, there are so-called interval multiplication truncated data. This poses a challenge to the accurate identification of the joint CDF.

[0003] The gradation of deep-sea manganese nodules refers to the particle size distribution of ore particles, which is crucial to the design and performance of mining robots. Accurate prediction and simulation of gradation can ensure that mining robots can effectively collect ore particles of different sizes in actual operations, improving collection efficiency and accuracy. However, these data often exist in the form of binary type I interval truncation, which requires the development of new statistical modeling methods to deal with these problems. Since the size and mass data of manganese nodules are strongly correlated, and this correlation has a direct impact on the picking efficiency of mining robots, therefore, establishing an accurate joint probability distribution model is crucial to optimize the design of mining robots.

[0004] At the same time, in the development of deep-sea mining technology, the design optimization of the ore collection device is a complex issue because it involves multiple variables and constraints. Improving the collection rate of nodule particles is a key indicator to measure the performance of the ore collection device. However, in order to achieve this goal, multiple design parameters need to be considered at the same time, such as the initial jet velocity, nozzle width, nozzle height from the bottom, etc. These parameters are interdependent and have a direct impact on the collection rate. Traditional optimization methods often focus on the adjustment of a single parameter, and it is difficult to take into account the optimization of multiple parameters at the same time, resulting in the failure to achieve the best collection effect.

[0005] In this context, it is particularly important to adopt a multi-parameter optimization method. By comprehensively considering various design parameters, the performance of the ore collection device can be more comprehensively understood and predicted, so as to achieve the optimization of the device design while ensuring the collection rate. This optimization method requires the algorithm to be able to handle the interactions and constraints between parameters and find the best combination of parameters to maximize the collection rate.

[0006] Traditional optimization methods often have difficulty handling this complexity, especially when faced with discrete variables and complex nonlinear relationships. Discrete genetic algorithms (DGAs) are an effective search and optimization tool that can handle optimization problems with discrete variables. They iteratively improve potential solutions by simulating natural selection and genetic mechanisms.

[0007] Therefore, the present invention aims to optimize the design of the ore collecting device and improve the efficiency of deep-sea mining by simulating the deep-sea seabed particle grading problem through a multi-parameter optimization design method of the ore collecting device based on a genetic algorithm under simulation conditions. Summary of the invention

[0008] The present invention provides a multi-parameter optimization design method for a ore-collecting device based on deep-sea nodule grading. The method adopts a discrete genetic algorithm to perform overall optimization on multiple design parameters of the ore-collecting device, introduces an elite retention strategy and a uniform two-point crossover method, and adopts a maximum likelihood method to estimate the parameters of the marginal probability density function of nodule size and mass, thereby constructing a size-mass joint probability distribution function of nodule particles. The method can more accurately fit the size and mass distribution of nodules, create a more realistic seabed environment model for the subsequent design of ore-collecting devices, and provide a more realistic distribution of seabed nodule particles, so as to solve the problems raised in the above-mentioned background technology.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A multi-parameter optimization design method for a deep-sea nodule grading-based ore collection device comprises the following steps:

[0011] S1: Data collection and preprocessing of nodule sea trial samples; nodules of different shapes are equivalent to spherical nodules of corresponding diameters, and the mass and particle size information of nodules of each sample are counted;

[0012] S2: likelihood function defining nodule size and mass;

[0013] S3: Estimate the parameters of the two marginal probability density functions using the maximum likelihood method (MLE);

[0014] S4: The most appropriate marginal distribution is selected from the candidate probability distributions using the maximum value of the log-likelihood function and the number of parameters through the Akaike Information Criterion (AIC);

[0015] S5: The copula function family is used to fit the joint probability distribution of size and mass. The size and mass of polymetallic nodules are highly correlated, and it is necessary to fit multiple variables to determine the particle grading parameters.

[0016] S6: Estimate the parameters of its joint probability density function using maximum likelihood;

[0017] S7: Using the AIC criterion, the most suitable joint distribution is selected from the candidate probability distributions to obtain the joint probability density function of size-mass;

[0018] S8: Prepare the seafloor particle data required for the simulation; use the particle generation tool of the simulation software to generate particles according to the size-mass joint distribution function; arrange the generated particles in the calculation domain to simulate the distribution of the seafloor particle layer; adjust the initial position and distribution of the particles to ensure that they conform to the distribution characteristics of the actual seafloor particles;

[0019] S9: Select the design parameters of the mining head based on the Coanda effect;

[0020] S10: converting the continuous design variables in the genetic algorithm GA into discrete design variables, i.e. using a discrete genetic algorithm (DGA) on the optimization of the design parameters of the Coanda effect set mine head; the discrete genetic algorithm uses a binary code string in the genetic algorithm to represent the position of each parameter value in the range of each parameter value, and performs selection, crossover and mutation operations on the position represented by the binary code string to obtain a new generation of chromosomes; at the same time, a uniform two-point crossover method and an elite retention strategy method are introduced;

[0021] S11: testing the discrete genetic algorithm through a standard test function; if the test result of the discrete genetic algorithm meets the preset requirements, implicitly optimizing the unsteady characteristics of the design parameters of the Coanda effect ore collection device through the discrete genetic algorithm after the test.

[0022] Preferably, the design parameters of the Coanda effect ore collection device include the following design parameters:

[0023] The curvature radius of the jet surface R;

[0024] Jet nozzle width h;

[0025] Height of jet nozzle from bottom H n ;

[0026] The height of the jet surface from the bottom is H;

[0027] Nozzle jet angle θ;

[0028] Initial jet velocity U i ;

[0029] Jet nozzle width h / jet surface curvature radius R.

[0030] Preferably, the implicit unsteady characteristic optimization of the Coanda effect ore collection device is specifically as follows:

[0031] (1) Determine the optimization target, and take the nodule particle collection rate of the Coanda effect collection device as the single optimization target;

[0032] (2) Determine the optimization parameters and calculation domain;

[0033] (3) Setting algorithm parameters;

[0034] (4) Using mathematical computing tools such as MATLAB, the Latin hypercube sampling method is used to initialize the chromosomes in the discrete computing domain to obtain the initial population;

[0035] (5) using the chromosome parameters obtained in step (4) to establish a Coanda effect mining device model corresponding to each chromosome through industrial modeling software such as UG;

[0036] (6) The Coanda effect ore collection device model obtained in step (5) is used to draw a ore collection device model grid through a grid division software such as ICEM;

[0037] (7) importing the ore collection device model grid obtained in step (6) into STAR-CCM, and arranging and generating particles in the model simulation domain according to the size-mass joint distribution function obtained in step S8, and performing implicit unsteady numerical simulation through STAR-CCM;

[0038] (8) Processing the data obtained in step (7) to obtain the particle collection rate of the ore collection device, and obtaining the fitness of the chromosome to the discrete calculation domain. The fitness here is a parameter related to the particle collection rate of the ore collection device. The higher the collection rate, the higher the fitness;

[0039] (9) Find the chromosome with the largest fitness value in the population, whose fitness is The corresponding decision variable values t is the current iteration number;

[0040] (10) Determine whether the number of iterations in step (9) reaches the maximum number of iterations. If so, stop the iteration and output the optimization results of each parameter. If not, go to step (5) to continue iterative optimization.

[0041] (11) The roulette wheel selection method is used to select individuals with higher fitness from some of the parent generations, and the elite retention strategy is used to retain the best individuals in the previous generation iteration to the next generation; the probability of the parent generation individual being selected in the roulette wheel is completely determined by the fitness;

[0042] (12) Select individuals according to the crossover probability and perform crossover operations on them two by two to generate new individuals;

[0043] (13) Select individuals according to the mutation probability and perform mutation operations on them to generate new individuals;

[0044] (14) If a local optimal solution is generated, that is, the particle collection rate of the model established by the optimal individual parameters of the offspring is less than 90%, the parent generation, offspring, and local solution are merged into a population to generate a new generation population, the number of iterations is increased by one, and the process returns to step (5) to continue iterating until the result satisfies the particle collection rate of the iterative result model higher than 90%, and the optimization result is output;

[0045] (15) The parameter values ​​of the optimized multiple optimal solutions are divided into smaller intervals, and the optimization is continued in step (2) until the convergence requirements are met.

[0046] Preferably, step (3) sets the algorithm parameters, including the population number n p , crossover rate p c , mutation rate p m , maximum number of iterations n iter , the number of discrete variables n and their corresponding values.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. By using discrete genetic algorithm (DGA) to optimize the overall design parameters of the ore collection device, the collection efficiency is improved.

[0049] 2. Introduce the elite retention strategy and uniform two-point crossover method in DGA. The elite retention strategy ensures that the best individuals in the parent population can be retained to the next generation, reducing the possibility of "falling into local optimality" in parameter optimization. Select several individuals with the highest fitness as elite individuals, and these individuals will be directly copied to the next generation population without crossover or mutation operations. The uniform two-point crossover method generates new individuals by randomly selecting two points (parameters) in the gene strings of two parent individuals and then exchanging the gene segments between the two points. This helps to explore new parameter combinations and increase the diversity of the population, which helps to find better design solutions.

[0050] 3. The maximum likelihood method (MLE) is used to estimate the parameters of the marginal probability density function of nodule size and mass, and the most appropriate marginal distribution is selected from the candidate probability distributions by the Akaike Information Criterion (AIC). In this way, the size and mass distribution of nodules can be fitted more accurately, creating a more realistic seabed environment model for the subsequent design of the ore collection device and providing a more realistic distribution of seabed nodule particles.

[0051] 4. Accurate identification of multivariate distributions. Define the likelihood function using multinomial distribution and select the most appropriate marginal distribution using the Akaike Information Criterion; then estimate the parameters of the selected marginal distribution using the Maximum Likelihood Estimation (MLE) to maximize the likelihood function of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a parameter optimization flow chart of the present invention. DETAILED DESCRIPTION

[0053] 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.

[0054] See also Figure 1 The present invention provides a multi-parameter optimization design method for a deep-sea nodule grading-based ore collection device, comprising the following steps:

[0055] S1: Data collection and preprocessing of nodule sea trial samples; nodules of different shapes are equivalent to spherical nodules of corresponding diameters, and the mass and particle size information of nodules of each sample are counted;

[0056] S2: Define the likelihood function of nodule size and mass. Assume that the marginal probability density functions of size and mass are f(x,θ1) and g(y,θ2) respectively. The likelihood function can be expressed as:

[0057]

[0058] In the formula, θ1, θ2 represent the parameter vectors of the marginal distribution of the variable, x i ,y i is the sample, n is the sample size;

[0059] S3: Use the maximum likelihood method (MLE) to estimate the parameters of the two marginal probability density functions. The maximum likelihood estimate can be expressed as:

[0060]

[0061]

[0062] The maximum likelihood method (MLE) is used to estimate the parameters of the marginal probability density function of nodule size and mass, and the most appropriate marginal distribution is selected from the candidate probability distributions by the Akaike Information Criterion (AIC). In this way, the size and mass distribution of nodules can be fitted more accurately, creating a more realistic seabed environment model for the subsequent design of the ore collection device and providing a more realistic distribution of seabed nodule particles;

[0063] S4: The most appropriate marginal distribution is selected from the candidate probability distributions using the maximum value of the log-likelihood function and the number of parameters through the Akaike Information Criterion (AIC); the log-likelihood function is defined as:

[0064]

[0065] Where f(x;θ) is the PDF value of the marginal distribution, θ is the parameter vector of the marginal distribution, and x i is the i-th reference point, m marginal is the number of given data;

[0066] The AIC criterion considers the maximum value of the log-likelihood function and the number of parameters and can be expressed as:

[0067] AIC=2k-2ln

[0068] Where k is the number of parameters of each candidate distribution, and ln is the maximum log-likelihood value. The likelihood function is defined by using a multinomial distribution, and the most appropriate marginal distribution is selected using the Akaike Information Criterion (AIC). The parameters of the selected marginal distribution are then estimated by maximum likelihood estimation (MLE) to maximize the likelihood function of the model.

[0069] S5: The copula function family is used to fit the joint probability distribution of size and mass. The size and mass of polymetallic nodules are highly correlated, and it is necessary to fit multiple variables to determine the particle grading parameters. The joint probability density function can be expressed as:

[0070]

[0071] Where F(x,θ1) and G(y,θ2) are the marginal distribution functions of the variables, c is the copula function, and α is the unknown parameter of the copula function;

[0072] S6: Use maximum likelihood to estimate the parameters of its joint probability density function, and its formula can be expressed as:

[0073]

[0074] S7: Use the AIC criterion to select the most suitable joint distribution from the candidate probability distributions to obtain the joint probability density function of size-mass; define the likelihood function of the Copula, and also use the multinomial distribution to process the data; then use AIC to select the most suitable Copula, that is, the Copula with the smallest AIC value; finally, use MLE to estimate the parameters of the selected Copula;

[0075] S8: Prepare the seafloor particle data required for simulation, including information such as particle size and mass; use the particle generation tool of the simulation software to generate particles according to the size-mass joint distribution function; arrange the generated particles in the calculation domain to simulate the distribution of the seafloor particle layer; adjust the initial position and distribution of the particles to ensure that they conform to the distribution characteristics of the actual seafloor particles;

[0076] S9: Select the design parameters of the mining head based on the Coanda effect;

[0077] S10: Convert the continuous design variables in the genetic algorithm GA into discrete design variables, that is, use a discrete genetic algorithm (DGA) on the optimization of the design parameters of the Coanda effect ore collector; use the discrete genetic algorithm (DGA) to optimize the multiple design parameters of the ore collector as a whole, thereby improving the collection efficiency; the discrete genetic algorithm uses a binary code string in the genetic algorithm to represent the position of each parameter value in the range of each parameter value, and selects, crosses, and mutates the position represented by the binary code string to obtain a new generation of chromosomes; at the same time, introduce the uniform two-point crossover method and the elite retention strategy method; introduce the elite retention strategy and the uniform two-point crossover method in DGA. The elite retention strategy ensures that the best individuals in the parent population can be retained to the next generation, reducing the possibility of "falling into local optimality" in parameter optimization. Select several individuals with the highest fitness as elite individuals, and these individuals will be directly copied to the next generation population without crossover or mutation operations; the uniform two-point crossover method generates new individuals by randomly selecting two points (parameters) in the gene strings of two parent individuals, and then exchanging the gene segments between the two points. This helps to explore new parameter combinations and increase the diversity of the population, which helps to find better design solutions;

[0078] S11: testing the discrete genetic algorithm through a standard test function; if the test result of the discrete genetic algorithm meets the preset requirements, implicitly optimizing the unsteady characteristics of the design parameters of the Coanda effect ore collection device through the discrete genetic algorithm after the test.

[0079] As a further embodiment, the design parameters of the Coanda effect ore collection device include but are not limited to the following design parameters:

[0080] The curvature radius of the jet surface R;

[0081] Jet nozzle width h;

[0082] Height of jet nozzle from bottom H n ;

[0083] The height of the jet surface from the bottom is H;

[0084] Nozzle jet angle θ;

[0085] Initial jet velocity U i ;

[0086] Jet nozzle width h / jet surface curvature radius R.

[0087] As a further embodiment, the implicit unsteady characteristic optimization of the Coanda effect ore collection device is specifically as follows:

[0088] (1) Determine the optimization target, and take the nodule particle collection rate of the Coanda effect collection device as the single optimization target;

[0089] (2) Determine the optimization parameters and calculation domain, and pre-select the design parameter range of the ore collection device, including the initial jet velocity, nozzle width, etc. Discretize the continuous design parameters and encode them into binary strings. For example, each possible value of the initial jet velocity corresponds to a unique binary code. Select the initial jet velocity U i , height of jet surface from bottom H, height of jet nozzle from bottom H n , jet nozzle width h / jet surface curvature radius R, nozzle jet angle θ; the value range of the above optimization parameters is the calculation domain; the calculation domain is specifically: U i ∈[10,12,14]m / s;H∈[0.03,0.06,0.09]m;H n ∈[0.06,0.09,0.12]m;h / R∈[0.02,0.04,0.06];θ∈[40,45,50]°;

[0090] (3) Set algorithm parameters, including: population number n p , crossover rate p c , mutation rate p m , maximum number of iterations n iter , the number of discrete variables n and their corresponding values;

[0091] (4) Using mathematical computing tools such as MATLAB, the Latin hypercube sampling method is used to initialize the chromosomes in the discrete computing domain. MATLAB is selected as the main computing platform, and the genetic algorithm function GA in its Global Optimization Toolbox is used; the population size is set to 50, indicating that there are 50 candidate solutions in each iteration, and the initial population InitialData = {case1, case2, case3, ..., case 49 ,case 50};

[0092] (5) using the chromosome parameters obtained in step (4) to establish a Coanda effect mining device model corresponding to each chromosome through industrial modeling software such as UG;

[0093] (6) The Coanda effect ore collection device model obtained in step (5) is used to draw a ore collection device model grid through a grid division software such as ICEM;

[0094] (7) importing the ore collection device model grid obtained in step (6) into STAR-CCM, and arranging and generating particles in the model simulation domain according to the size-mass joint distribution function obtained in step S8, and performing implicit unsteady numerical simulation through STAR-CCM;

[0095] (8) The data obtained in step (7) is processed to obtain the particle collection rate of the ore collection device, and the fitness of the chromosome to the discrete calculation domain is obtained f = {f1, f2, f3, ... f 49 ,f 50}, fitness here is a parameter related to the particle collection rate of the ore collection device. The higher the collection rate, the higher the fitness. A fitness function is created, which is a weighted function of the collection rate η. The function evaluates the performance of each candidate solution according to the particle collection rate of the ore collection device;

[0096] (9) Find the chromosome with the largest fitness value in the population, whose fitness is The corresponding decision variable values t is the current iteration number;

[0097] (10) Set the maximum number of iterations to 100, set the termination condition of the optimization to reach the maximum number of iterations, extract the individual with the highest fitness from the final population as the optimal design parameter, and determine whether the number of iterations in step (9) has reached 100. If so, stop the iteration and output the optimization results of each parameter; if not, go to step (5) to continue iterative optimization;

[0098] (11) The roulette wheel selection method is used to select individuals with higher fitness from some of the parent generations, and the elite retention strategy is used to retain the best individuals in the previous iteration to the next generation. The probability of a parent individual being selected in the roulette wheel is completely determined by its fitness. If the fitness of a parent individual is f i , then the probability of being selected is P i for:

[0099]

[0100] (12) Select individuals according to the crossover probability and perform crossover operations on them two by two to generate new individuals;

[0101] (13) According to the mutation probability, individuals are selected and mutated to generate new individuals; in MATLAB, the crossover function is set to two-point crossover and the mutation function is set to uniform mutation. In the ga function, the crossover probability is set to 0.7 and the mutation probability is set to 0.04.

[0102] (14) If a local optimal solution is generated, that is, the particle collection rate of the model established by the optimal individual parameters of the offspring is less than 90%, the parent generation, offspring, and local solution are merged into a population to generate a new generation population, the number of iterations is increased by one, and the process returns to step (5) to continue iterating until the result satisfies the particle collection rate of the iterative result model higher than 90%, and the optimization result is output;

[0103] (15) The parameter values ​​of the optimized multiple optimal solutions are divided into smaller intervals, and the optimization is continued in step (2) until the convergence requirements are met.

[0104] As a further embodiment, step (3) setting algorithm parameters includes: p =50, crossover rate p c =0.7, mutation rate p m =0.04, maximum number of iterations n iter =100;

[0105] The standard test functions used are:

[0106]

[0107] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-parameter optimization design method for a collection device based on deep-sea nodule grading, characterized in that: The steps include: S1: Data collection and preprocessing of nodule sea trial samples; nodules of different shapes are equivalent to spherical nodules of corresponding diameters, and the mass and particle size information of nodules of each sample are counted; S2: likelihood function defining nodule size and mass; S3: Estimate the parameters of the two marginal probability density functions using the maximum likelihood method (MLE); S4: The most appropriate marginal distribution is selected from the candidate probability distributions using the maximum value of the log-likelihood function and the number of parameters through the Akaike Information Criterion (AIC); S5: The copula function family is used to fit the joint probability distribution of size and mass. The size and mass of polymetallic nodules are highly correlated, and it is necessary to fit multiple variables to determine the particle grading parameters. S6: Estimate the parameters of its joint probability density function using maximum likelihood; S7: Using the AIC criterion, the most suitable joint distribution is selected from the candidate probability distributions to obtain the joint probability density function of size-mass; S8: Prepare the seafloor particle data required for the simulation; use the particle generation tool of the simulation software to generate particles according to the size-mass joint distribution function; arrange the generated particles in the calculation domain to simulate the distribution of the seafloor particle layer; adjust the initial position and distribution of the particles to ensure that they conform to the distribution characteristics of the actual seafloor particles; S9: Select the design parameters of the mining head based on the Coanda effect; S10: converting the continuous design variables in the genetic algorithm GA into discrete design variables, i.e. using a discrete genetic algorithm (DGA) on the optimization of the design parameters of the Coanda effect set mine head; the discrete genetic algorithm uses a binary code string in the genetic algorithm to represent the position of each parameter value in the range of each parameter value, and performs selection, crossover and mutation operations on the position represented by the binary code string to obtain a new generation of chromosomes; at the same time, a uniform two-point crossover method and an elite retention strategy method are introduced; S11: testing the discrete genetic algorithm through a standard test function; if the test result of the discrete genetic algorithm meets the preset requirements, implicitly optimizing the unsteady characteristics of the design parameters of the Coanda effect ore collection device through the discrete genetic algorithm after the test.

2. The multi-parameter optimization design method for a deep-sea nodule grading-based ore collection device according to claim 1, characterized in that: The design parameters of the Coanda effect ore collection device include the following design parameters: The curvature radius of the jet surface R; Jet nozzle width h; Height of jet nozzle from bottom H n ; The height of the jet surface from the bottom is H; Nozzle jet angle θ; Initial jet velocity U i ; Jet nozzle width h / jet surface curvature radius R.

3. The multi-parameter optimization design method for a deep-sea nodule grading-based ore collection device according to claim 1, characterized in that: The implicit unsteady characteristic optimization of the Coanda effect ore collection device is specifically as follows: (1) Determine the optimization target, and take the nodule particle collection rate of the Coanda effect collection device as the single optimization target; (2) Determine the optimization parameters and calculation domain; (3) Setting algorithm parameters; (4) Using mathematical computing tools such as MATLAB, the Latin hypercube sampling method is used to initialize the chromosomes in the discrete computing domain to obtain the initial population; (5) using the chromosome parameters obtained in step (4) to establish a Coanda effect mining device model corresponding to each chromosome through industrial modeling software such as UG; (6) The Coanda effect ore collection device model obtained in step (5) is used to draw a ore collection device model grid through a grid division software such as ICEM; (7) importing the ore collection device model grid obtained in step (6) into STAR-CCM, and arranging and generating particles in the model simulation domain according to the size-mass joint distribution function obtained in step S8, and performing implicit unsteady numerical simulation through STAR-CCM; (8) Processing the data obtained in step (7) to obtain the particle collection rate of the ore collection device, and obtaining the fitness of the chromosome to the discrete calculation domain. The fitness here is a parameter related to the particle collection rate of the ore collection device. The higher the collection rate, the higher the fitness; (9) Find the chromosome with the largest fitness value in the population, whose fitness is The corresponding decision variable values t is the current iteration number; (10) Determine whether the number of iterations in step (9) reaches the maximum number of iterations. If so, stop the iteration and output the optimization results of each parameter. If not, go to step (5) to continue iterative optimization. (11) The roulette wheel selection method is used to select individuals with higher fitness from some of the parent generations, and the elite retention strategy is used to retain the best individuals in the previous generation iteration to the next generation; the probability of the parent generation individual being selected in the roulette wheel is completely determined by the fitness; (12) Select individuals according to the crossover probability and perform crossover operations on them two by two to generate new individuals; (13) Select individuals according to the mutation probability and perform mutation operations on them to generate new individuals; (14) If a local optimal solution is generated, that is, the particle collection rate of the model established by the optimal individual parameters of the offspring is less than 90%, the parent generation, offspring, and local solution are merged into a population to generate a new generation population, the number of iterations is increased by one, and the process returns to step (5) to continue iterating until the result satisfies the particle collection rate of the iterative result model higher than 90%, and the optimization result is output; (15) The parameter values ​​of the optimized multiple optimal solutions are divided into smaller intervals, and the optimization is continued in step (2) until the convergence requirements are met.

4. The multi-parameter optimization design method for a ore collection device based on deep-sea nodule grading according to claim 3 is characterized in that: Step (3) Set the algorithm parameters, including the population number n p , crossover rate p c , mutation rate p m , maximum number of iterations n iter , the number of discrete variables n and their corresponding values.

Images