Multi-mode laser data management system and method based on fuzzy neural network

Abstract

The invention belongs to the technical field of data management, and discloses a multi-mode laser data management system and method based on a fuzzy neural network. Comprising the following steps: collecting laser experiment data, an experiment base material type and an experiment reliability evaluation degree; carrying out normalization processing on the laser experiment data; fuzzy calculation of the membership degree is carried out on the normalized laser experiment data based on a membership degree function; training a machine learning model for generating a prediction missing experiment parameter membership degree based on the membership degree, the experiment base material type and the experiment reliability evaluation degree; based on the input and output of a machine learning model, converting into a fuzzy rule; taking a weighted average membership value in a fuzzy rule reasoning result as the output of fuzzy synthesis; defuzzification is carried out through a gravity center method based on output of fuzzy synthesis, and laser experiment data are filled with experiment parameters after defuzzification and marked as intact experiment data; and data analysis is carried out by using intact experimental data.

Claims

1. A multimodal laser data management method based on fuzzy neural network, characterized in that: Including: Collecting laser experiment data, the type of experimental substrate, and the reliable evaluation degree of the experiment; the laser experiment data includes five experimental parameters: laser power, scanning speed, scanning radius, laser frequency, and laser distance. Normalizing the laser experiment data; dividing different types of experimental parameters into 5 fuzzy levels respectively; each fuzzy level corresponds to a fuzzy set. Performing fuzzy calculation of membership degrees on the normalized laser experiment data based on the membership function. Training a machine learning model for generating the membership degree of the missing experimental parameter based on the membership degree, the type of experimental substrate, and the reliable evaluation degree of the experiment. Based on the input and output of the machine learning model, converting them into fuzzy rules; the premise part of the fuzzy rule is the membership degree in the input of the machine learning model, and the conclusion part is the output of the machine learning model; taking the fuzzy level with the highest membership degree as the fuzzy level to which the experimental parameter belongs. Performing fuzzy composition based on the fuzzy rules. Taking the weighted average membership degree value in the fuzzy rule inference result as the output of the fuzzy composition. Performing defuzzification on the output of the fuzzy composition by the centroid method, and filling the defuzzified experimental parameter back into the laser experiment data and marking it as complete experimental data; using the complete experimental data for data analysis.

2. The multimodal laser data management method based on fuzzy neural network according to claim 1, characterized in that: The acquisition method of the laser experiment data includes identifying relevant laser cleaning literature through OCR technology to obtain the published laser experiment data. The normalization process includes: Calculate the normalized value; where x′ is the normalized value, ranging from [0,1]; x is the original experimental parameter; x max is the maximum value of the experimental parameter; x min is the minimum value of the experimental parameter; Normalizing different experimental parameters (laser power, scanning speed, scanning radius, laser frequency, and laser distance) respectively; presetting 4 representative values for each fuzzy set, which are a, b, c, and d; the fuzzy levels include: low, sub-low, medium, sub-high, and high.

3. The multimodal laser data management method based on fuzzy neural network according to claim 2 is characterized in that: The membership degree formula is as follows: In the formula, a, b, c, and d are all numerical points in the fuzzy set, and a < b < c < d; a corresponds to the starting point of the left boundary of the fuzzy set, and d corresponds to the ending point of the right boundary of the fuzzy set; in this formula, x is the normalized laser experiment data; μ(x) is the membership degree value of the laser experiment data; the range of the membership degree value is [0, 1]. The experimental substrate type is the type of material cleaned by laser; the experimental reliability evaluation degree R = 0.4J 2 +10I+T; where J is the preset journal grade evaluation value; I is the journal impact factor; T is the preset journal author professional title grade evaluation value.

4. The multimodal laser data management method based on fuzzy neural network according to claim 3 is characterized in that: The training process of the machine learning model includes: Listing the membership degree, the type of experimental substrate, and the reliable evaluation degree of the experiment in the same group as a group of first-level training data; collecting four consecutive groups of training data as second-level training data; the label of a group of second-level training data is the membership degree of the missing experimental parameter in the fifth group of first-level training data; taking a group of second-level training data and its label as a group of samples, and collecting multiple groups of samples as a data set; the data set is divided into a training set, a validation set, and a test set, where the training set accounts for 60% of the data set, and the validation set and the test set each account for 20% of the data set. Taking the training set as the input of the machine learning model, and the machine learning model takes the membership degree of the missing experimental parameter in the fifth group of first-level training data as the output; taking the label corresponding to a real-time group of second-level training data as the prediction target, and taking minimizing the loss function value of the machine learning model as the training target; stopping training when the loss function value of the machine learning model is less than or equal to the preset target loss value. The machine learning model loss function is the mean square error; the mean square error is calculated by Minimize to train the model; in the loss function, MSE is the loss function value, i is the secondary training data group number; u is the number of secondary training data groups; y i is the label corresponding to the i-th group of secondary training data, The membership degree of the missing experimental parameters of the fifth group of primary training data predicted by the i-th group of secondary training data; The machine learning model is a deep neural network model. The calculation method of the hidden layer H of the deep neural network model includes: H=G[b i +w i s(b i-1 +w i-1 x)]; where H is the output of the hidden layer; G is the activation function; b i is the bias term of the hidden layer; w i is the weight of the hidden layer; s is the activation function; b i-1 is the bias term of the previous layer; w i-1 is the weight of the previous layer; x is the input training set.

5. The multimodal laser data management method based on fuzzy neural network according to claim 4, characterized in that: The defuzzification method is the centroid method, and the formula is Where μ(x) is the membership value of fuzzy synthesis, x is the representative value (a, b, c and d) in the fuzzy set; * are the experimental parameters for defuzzification.

6. A multimodal laser data management system based on fuzzy neural network, characterized in that: Including: Data collection module, which collects laser experimental data, experimental substrate type and experimental reliability evaluation; laser experimental data includes laser power, scanning speed, scanning radius, laser frequency and laser distance, a total of five experimental parameters; The preprocessing module normalizes the laser experimental data; each of the different types of experimental parameters is divided into five fuzzy levels; each fuzzy level corresponds to a fuzzy set; The fuzzification module performs fuzzification calculation of the membership of the normalized laser experimental data based on the membership function; Model training module, which trains a machine learning model to generate membership of predicted missing experimental parameters based on membership, experimental substrate type and experimental reliability assessment; The fuzzy reasoning module converts the input and output of the machine learning model into fuzzy rules; the premise of the fuzzy rule is the membership degree in the machine learning model input, and the conclusion is the output of the machine learning model; the fuzzy level with the highest membership degree is taken as the fuzzy level to which the experimental parameter belongs; The fuzzy synthesis module performs fuzzy synthesis based on fuzzy rules; the weighted average membership value in the fuzzy rule reasoning result is used as the output of the fuzzy synthesis; The fuzzy output module defuzzifies the output based on the fuzzy synthesis through the centroid method, and fills the defuzzified experimental parameters back into the laser experimental data and marks them as intact experimental data; the intact experimental data is used for data analysis.

7. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the multimodal laser data management method based on fuzzy neural network as described in any one of claims 1 to 5 by calling the computer program stored in the memory.

8. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer is caused to execute a multi-modal laser data management method based on a fuzzy neural network as claimed in any one of claims 1 to 5.

Images