Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Remaining Life Prediction Method for Monotone Echo State Networks

An echo state network and life prediction technology, applied in special data processing applications, instruments, electrical digital data processing, etc., can solve problems such as low computing efficiency, poor approximation performance, and inability to accurately predict lifespan, etc., to improve approximation performance, improve The effect of approximating efficiency and narrowing the scope of optimization

Active Publication Date: 2016-01-20
HARBIN INST OF TECH
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problem of low calculation efficiency, poor approximation performance and other shortcomings in the approximation of functions with monotonic trends in traditional ESNs, and the problem that life cannot be accurately predicted, and provides a remaining life prediction for monotone echo state networks method

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Remaining Life Prediction Method for Monotone Echo State Networks
  • Remaining Life Prediction Method for Monotone Echo State Networks
  • Remaining Life Prediction Method for Monotone Echo State Networks

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0028] Specific implementation mode one: the following combination figure 2 Describe this embodiment, the method for predicting the remaining life of the monotone echo state network described in this embodiment, step 1, randomly establish the monotone echo state network model MONESN:

[0029] The monotone echo state network model MONESN includes an input unit u(n), an internal processing unit x(n) and an output unit y(n), where n is the moment of the corresponding system state transition; the input unit collects simulated physical parameters of the turbine engine, including temperature , pressure and speed parameters of the system; or the input unit collects the charge and discharge data of the lithium-ion battery;

[0030] u(n) represents the value of the input unit at time n, x(n) represents the value of the internal processing unit at time n, and y(n) represents the value of the output unit at time n.

[0031] The output unit outputs the number of remaining operating cycl...

specific Embodiment approach 2

[0039] Specific implementation mode two: this implementation mode further explains implementation mode one, and the monotone echo state network model MONESN randomly established in step one includes an input unit u(n), an internal processing unit x(n) and an output unit y(n),

[0040] u(n)=(u 1 (n), u 1 (n),...,u j (n),...,u L (n)),j=1,2,...,L;

[0041] x(n)=(x 1 (n), x 2 (n),...,x s (n),...,x N (n)), s=1,2,...,N;

[0042] y(n)=(y 1 (n),y 2 (n),...,y i (n),...,y M (n)), i=1,2,...,M;

[0043] n is the moment corresponding to the state transition of the system, L, M and N are all positive integers;

[0044] The process of establishing MONESN is:

[0045] Step 11. Randomly establish an N×L-dimensional input weight matrix W in , N×N-dimensional internal connection weight matrix W 0 and N×M dimensional feedback weight matrix W back ;

[0046] Step 12, according to the formula:

[0047] W 1 =W 0 / |λ max |

[0048] get W 1 , where |λ max | is W 0 The absolut...

specific Embodiment approach 3

[0054] Specific implementation mode three: this implementation mode further explains implementation mode one, the process of carrying out network dynamic training to the untrained MONESN that step one is set up described in step two is:

[0055] Step 21, initialize the internal processing unit x(0)=0, output unit y(0)=0;

[0056] Step 22, input sequence u j (n) and the real output sample sequence y i (n) Drive MONESN and update the equation according to the internal processing unit of MONESN

[0057] x(n)=f(W in u(n)+Wx(n-1)+W back y(n-1))

[0058] and

[0059] y(n)=W out (u(n),x(n))

[0060] Get the internal neuron state at each moment;

[0061] Store the input and internal state at each moment in the form of row vectors in the T×(L+N)-dimensional internal state matrix C; and store the real output at the corresponding moment in the form of row vectors in T×M as the output matrix d;

[0062] Step two and three, constructing constraint inequality: the transposition of...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a method for predicating residual life by a monotonous echo state network (MONESN), belongs to the field of failure prediction and system health management, and aims to solve the problem that the traditional ESN cannot accurately predict life. The method comprises the following steps: (1) randomly establishing an MONESN model; (2) performing network dynamic training to obtain the monotonically increased or monotonically decreased output weight of the MONESN model, and substituting the output weight into the MONESN to obtain a trained MONESN; and (3) outputting the residual operational cycle life of a turbine engine to be subjected to life prediction to an input unit of the trained MONESN in the step (2), and outputting the residual life of the turbine engine to be subjected to life predication by using the MONESN, or outputting the residual operational cycle life of a lithium ion battery to be subjected to life prediction to the input unit of the trained MONESN in the step (2), and outputting the residual life of the lithium ion battery to be subjected to life predication by using the MONESN.

Description

technical field [0001] The invention relates to a method for predicting the remaining life of a monotone echo state network, and belongs to the field of fault prediction and system health management. Background technique [0002] The function approximation problem is a basic problem in neural network research, which mainly depends on the highly nonlinear approximation ability of neural network. It has been shown that feedforward networks can approximate measurable functions on arbitrary compact domains (Borel-fields). The recurrent neural network can be understood as an open dynamic system, which describes the internal mapping of the recurrent neural network in the form of state space, as shown in formula 1: [0003] the s t+1 =f(As t +Bu t +θ)state transition [0004] the y t =Cs t output equation(1) [0005] Among them, A, B, C are the weight matrix, θ is the deviation unit, and is the input variable u t bias. f is the activation function of the neural network, u...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F19/00
Inventor 彭宇刘大同王红郭力萌彭喜元
Owner HARBIN INST OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com