A phase modulator pipeline data completion method with automatic optimization parameter tuning capability

Abstract

The application discloses a phase modulator pipeline data completion method with automatic optimization parameter adjustment capability, comprising the following steps: collecting data parameters of a phase modulator pipeline; importing the data parameters into an RBF neural network model for training to obtain optimization parameters; importing the optimization parameters into a crossover algorithm for parameter vector calculation to obtain average errors of all arrays of the optimization parameters; evaluating the parameter vectors based on the average errors of all arrays of the optimization parameters, and respectively defining fitness of the parameter vectors corresponding to all arrays of the optimization parameters; importing the fitness of the parameter vectors corresponding to all arrays into a genetic algorithm for global optimization, and outputting the optimization parameters whose fitness is higher than that of other parameter vectors. The application trains optimization parameters by using an RBF neural network model, verifies the fitness of the optimization parameters by combining a crossover algorithm, determines an optimal parameter vector according to a genetic algorithm, and then obtains optimal parameters to model, so that pipeline completion data can be predicted with high precision by a network model, and the error of the completion data is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of intelligent sensing of phase modulation machine pipelines, and particularly relates to a phase modulation machine pipeline data completion method with automatic optimization parameter adjustment capability. BACKGROUND

[0002] In actual operation and maintenance of the phase modulation machine in the battery swap station, it is found that the operation state of the pipeline matched with the phase modulation machine will affect the working condition of the phase modulation machine, and when the pipeline has serious vibration anomaly, the phase modulation machine itself will even be damaged. However, when the phase modulation machine pipeline is detected, due to the cost, sensors cannot be arranged at all positions of the pipeline, and at some narrow positions and dangerous positions, maintenance workers cannot conveniently and safely install sensors. Therefore, the limited vibration data obtained are needed to predict the pipeline nodes that are not used.

[0003] With the development and progress of deep learning theory and artificial intelligence technology, many intelligent pattern recognition methods can be used to predict the vibration characteristics of the pipeline, but most deep learning technologies depend on a large number of sample points, which do not conform to the operation characteristics of the phase modulation machine pipeline, and lack a corresponding hyperparameter selection method.

[0004] The radial basis neural network method can be used to complete the missing data of the pipeline characteristics using a small number of samples, but directly using the measured and simulated data of the phase modulation machine will cause a too large prediction error and cannot be used. SUMMARY

[0005] The application aims at overcoming the deficiencies in the prior art, and provides a phase modulation machine pipeline data completion method with automatic optimization parameter adjustment capability. The measured data and simulated data of the phase modulation machine pipeline are imported into an RBF neural network for parameter optimization, the fitness of a parameter vector is verified by using a crossover algorithm, the fitness of the parameter vector is globally optimized according to a genetic algorithm, the optimal parameter vector is determined, and the missing pipeline parameters are predicted and completed based on the parameter, so as to reduce the error of the completed data and provide a reference for vibration maintenance and treatment of the phase modulation machine pipeline.

[0006] Technical scheme: In a first aspect, the application provides a phase modulation machine pipeline data completion method with automatic optimization parameter adjustment capability, which comprises the following steps.

[0007] Collecting data parameters of the phase modulation machine pipeline;

[0008] Importing the data parameters into an RBF neural network model for training to obtain optimized parameters;

[0009] Importing the optimized parameters into a crossover algorithm for parameter vector calculation to obtain the average error of all arrays of the optimized parameters;

[0010] The parameter vector is evaluated based on the average error of all arrays, and the fitness corresponding to the parameter vector of all arrays of optimization parameters is defined respectively;

[0011] The fitness corresponding to the parameter vector of all arrays is imported into the genetic algorithm for global optimization, and the parameter vector with higher fitness than others is output;

[0012] The optimization parameter is determined according to the parameter vector with higher fitness than others, wherein the optimization parameter is used to import the RBF neural network to complete the prediction of the data, and the predicted completed data is output.

[0013] In further embodiments, the data parameters include: measured data and simulation data.

[0014] In further embodiments, the measured data is vibration data of the abnormal pipeline collected by a sensor.

[0015] The simulation data is converted, calculated and arranged to form a simulation database.

[0016] In further embodiments, the simulation data is converted, calculated and arranged to form a simulation database, and the method for converting, calculating and arranging the simulation data is to convert, calculate and arrange vibration data of different sources and forms.

[0017] The conversion and calculation include conversion of data including the vibration amplitude A of the pipeline, the vibration frequency f, and the pipeline vibration velocity v.

[0018] The pipeline vibration velocity is derived from the pipeline vibration displacement, and the expression of the pipeline vibration displacement is:

[0019] L = Asin (ωt + φ) (1)

[0020] The calculation formula of the pipeline vibration velocity v is:

[0021] v = -Aωcos (ωt + φ) (2)

[0022] The calculation formula of the maximum velocity of the vibration is:

[0023] v max = |Aω| (3)

[0024] In the formula, ω is the angular velocity of the pipeline, A and -A are the displacement in the vibration direction, and φ is the initial phase of the pipeline vibration.

[0025] In further embodiments, the RBF neural network model adjusts the structure of the prediction model according to different RBF neuron numbers n and diffusion factors S.

[0026] In further embodiments, the parameter vector is evaluated based on the average error of all arrays, and the fitness of the parameter vector corresponding to all arrays of optimization parameters is defined respectively, wherein the expression for defining the fitness is:

[0027]

[0028] In the formula, m is the total number of data, y i is the predicted value after the i-th modeling, y i is the corresponding measured value, and P(S, n) is the corresponding fitness.

[0029] In further embodiments, the genetic algorithm performs global optimization on the optimization parameters through a preset chromosome population number and total evolution number.

[0030] Advantages: Compared with the prior art, the present application has the following advantages:

[0031] The measured data and simulation data of the phase modifier pipeline collected are imported into the RBF neural network for parameter optimization, and the fitness of the parameter vector is verified by using the crossover algorithm, the fitness of the parameter vector is globally optimized according to the genetic algorithm, the optimal parameter vector is determined, and the RBF neural network is modeled based on the use of the vector, and data completion can be performed, which can ensure a small prediction error, predict and complete the missing pipeline parameters, and reduce the error of the completed data, thereby providing a reference for the vibration maintenance and management of the phase modifier pipeline. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The flowchart of the phase modifier pipeline data completion method with automatic optimization of the present application;

[0033] Figure 2 The logic diagram of the crossover algorithm and fitness solution of the present application;

[0034] Figure 3 The example diagram of the phase modifier pipeline of the present application.

[0035] DETAILED DESCRIPTION (Determine the technical solution after writing the specific embodiment)

[0036] In order to more fully understand the technical content of the present application, the technical solutions of the present application are further introduced and explained below in combination with specific embodiments, but are not limited thereto.

[0037] In combination with Figures 1 to 3 Further illustrate a phase modifier pipeline data completion method with automatic optimization of the present application in the embodiment, the specific steps are as follows:

[0038] Step 1: collect data parameters of the phase modifier pipeline; wherein the data parameters include: measured data and simulation data; the measured data is vibration data of the abnormal pipeline collected by a sensor; reference Figure 1 In this embodiment, the vibration data of the phase modifier abnormal pipeline is obtained using a sensor, but due to the cost of the sensor and the constraint of the pipeline structure, the data of the position of measuring point 3 is short, and the vibration speed data V1 of the remaining positions is used to predict the data sample of measuring point 3, which provides a reference for subsequent pipeline management. The simulation data is the conversion calculation and arrangement of the vibration data to form a simulation database;

[0039] Step 2: import the data parameters into the RBF neural network model for training to obtain the optimized parameters; reference Figure 1 In this embodiment, the simulation database is used to train the RBF neural network through a Matlab program; by controlling the key parameters of the network, the number of neurons n and the diffusion factor S, the structure of the model can be adjusted. Assuming that the prediction value after the i-th modeling is y i , the corresponding measured value is y i , E is the error of cross-validation, and P(S, n) is the corresponding fitness.

[0040] Step 3: import the optimized parameters into the cross algorithm for parameter vector calculation to obtain the average error of all arrays of the optimized parameters; reference Figure 2 In this embodiment, the error and fitness of each [S, n] combination are calculated using the cross-validation method.

[0041] Step 4: evaluate the parameter vector based on the average error of all arrays, and define the fitness of the parameter vector corresponding to all arrays of the optimized parameters respectively;

[0042] Step 5: import the fitness of all arrays of the parameter vector into the genetic algorithm for global optimization, and output the parameter vector with higher fitness than others;

[0043] Step 6: determine the optimized parameters according to the parameter vector with higher fitness than others, wherein the optimized parameters are used to import the RBF neural network for data completion prediction, and the predicted completed data is output; reference Figure 1 is made, the [S, n] combination that minimizes the validation error is obtained based on the genetic algorithm, and in the embodiment, the model error is reduced from 14 to 5.9; the cross probability pc is 0.6, the partial mapping crossover method is used, and the information of two chromosomes is exchanged; the mutation probability pm is 0.1, and the parameters in the chromosome are randomly modified within the value range; the upper and lower limits of the diffusion factor and the number of neurons are:

[0044]

[0045] The value after optimization is n=56, and S=1.62; therefore, the genetic algorithm performs global optimization on the optimization parameters by the preset chromosome population number and total evolution number.

[0046] Preferably, the simulation data is obtained by converting, calculating and arranging the vibration data, and the method for forming the simulation database is converting, calculating and arranging the vibration data of different sources and forms;

[0047] The conversion and calculation include converting the data containing the vibration amplitude A, vibration frequency f and pipeline vibration velocity v of the pipeline;

[0048] The pipeline vibration velocity is derived from the pipeline vibration displacement, and the expression of the pipeline vibration displacement is:

[0049] L=Asin (ωt+φ) (1)

[0050] The calculation formula of the pipeline vibration velocity v is:

[0051] v=-Aωcos (ωt+φ) (2)

[0052] The calculation formula of the maximum velocity of the vibration is:

[0053] v max =|Aω| (3)

[0054] In the formula, ω is the vibration angular velocity of the pipeline, A and -A are the displacement in the vibration direction, is the initial phase of the vibration of the pipeline.

[0055] Preferably, the parameter vector is evaluated based on the average error of all arrays, and the fitness of the optimization parameter array corresponding to the parameter vector is defined, and the expression of the fitness is defined as:

[0056]

[0057] In the formula, m is the total number of data, y i is the predicted value after the i-th modeling, y i is the corresponding measured value, and P(S, n) is the corresponding fitness.

[0058] As described above, the measured data and simulation data of the pipeline of the phase modifier are imported into the RBF neural network for parameter optimization, the fitness of the parameter vector is verified by using the crossover algorithm, the fitness of the parameter vector is globally optimized according to the genetic algorithm, the optimal parameter vector is determined, and the missing pipeline parameters are predicted and completed based on the parameters, so that the error of the completed data is reduced, and the vibration maintenance and treatment of the pipeline of the phase modifier are provided with reference.

[0059] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0060] The present application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0061] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0062] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0063] The above description is only preferred embodiments of the application. It should be pointed out that for those skilled in the art, some improvements and modifications can be made without departing from the technical principles of the application, and these improvements and modifications should also be considered as falling within the scope of the application.

Claims

1. A phase modulator pipeline data completion method with automatic optimization parameter tuning capability, characterized in that, The method comprises the following steps: Collecting data parameters of the phase modulation tube pipeline; Importing the data parameters into an RBF neural network model for training to obtain optimized parameters; Importing the optimized parameters into a crossover algorithm for parameter vector calculation to obtain the average error of all arrays of the optimized parameters; Evaluating the parameter vectors based on the average error of all arrays, and defining the fitness of the parameter vectors corresponding to all arrays of the optimized parameters, respectively; Importing the fitness of the parameter vectors corresponding to all arrays into a genetic algorithm for global optimization to output parameter vectors with higher fitness than others; Determining the optimized parameters according to the parameter vectors with higher fitness than others, wherein the optimized parameters are used to import the RBF neural network for prediction of the completed data, and the predicted completed data is outputted; The data parameters comprise measured data and simulation data; The measured data is vibration data of an abnormal pipeline collected by a sensor; The simulation data is obtained by converting, calculating and arranging the vibration data to form a simulation database; The method for converting, calculating and arranging the vibration data to form a simulation database comprises converting, calculating and arranging vibration data of different sources and forms; The conversion and calculation comprise converting data comprising the vibration amplitude A, the vibration frequency f and the pipeline vibration velocity v of the pipeline; The pipeline vibration velocity is derived from the pipeline vibration displacement, and the expression of the pipeline vibration displacement is: （1） The calculation formula of the pipeline vibration velocity v is: （2） The maximum velocity of the vibration is: （3）； where ω is the angular velocity of the pipe, A and are the displacements in the vibration directions, and φ is the initial phase of the pipe vibration.

2. The phase modulator tube data completion method with automatic optimization parameter adjustment capability according to claim 1, characterized in that, The RBF neural network model adjusts the structure of the prediction model according to different numbers n of RBF neurons and diffusion factors S.

3. The phase modulator tube data completion method with automatic optimization parameter adjustment capability according to claim 1, characterized in that, The parameter vectors are evaluated based on the average error of all arrays, and the fitness of the parameter vectors corresponding to all arrays of the optimized parameters is defined, respectively, wherein the expression of the fitness is: （4）； where m is the total number of data, is the predicted value after the i-th modeling, is the corresponding measured value, and P(S, n) is the corresponding fitness.

4. The phase modulator tube data completion method with automatic optimization parameter adjustment capability according to claim 1, characterized in that, The genetic algorithm performs global optimization on the optimized parameters by presetting the number of chromosome populations and the total number of evolution generations.

Images