Original link: http://tecdat.cn/?p=6829
Foreword
This paper presents a statistical method, using Bayesian inference impact position and impact force history identifier reinforced composite panel, which explicitly include uncertainty from modeling error and measurement noise. Represents the impact load by using a set of parameters, first converts the spatial domain (impact position) and time domain (impact force history) of the impact load parameter identification problem is to identify problems. Markov Chain Monte Carlo methods for sampling the posterior distribution to estimate the impact parameters. Noise using finite element numerical simulation of the data to demonstrate the effectiveness of the approach.
Brief introduction
In the aerospace industry, composite materials have been widely used in commercial and military components of the main structural load of the vehicle. One of the main problems is the design of the composite structure internal damage caused by low velocity impact, mainly layered, these damage difficult to detect and can significantly reduce the structural integrity. The latest developments and advances in computing and communications sensor technology allows people to research and development of health monitoring structures had a keen interest in these technologies can be integrated into the composite structure as built-in diagnostics system. For composite structures, the primary task to assess accurately the extent of damage and residual strength, efficient and reliable health monitoring system is to detect and identify the impact load when the collision event occurs.
Bayesian methods for identifying an impact load
In the present study, to impact loads coupled to a Bayesian framework identification, the first step is to use a set of parameters to represent the impact load.
Figure 1. The composite structure of an impact force history approximated
MCMC method
As a powerful stochastic simulation techniques, Monte Carlo (MC) method has been widely used to study the problems associated with probability. It can be very efficient, especially when the sample may be generated independently. Unfortunately, the posterior distribution of Bayesian inference used is usually very complex, it is difficult to draw independent samples for the standard MC method. In this case, MCMC Simulation is often used as an alternative to sampling. MCMC result is dependent on sample sequence (Markov chain), which has a stationary distribution equal to the target profile.
Numerical Simulation of matlab
Compare models forward collision
To demonstrate the effectiveness of the impact load identification of the proposed method, this section will study value. First, the former compared to the collision model using the finite element method and the method given in Section 4. For convenience, hereinafter, using a finite element method forward shock model called finite element model, the model used in this study is referred to as a forward shock model.
Numerical Study reinforcing composite panel and sensor placement
Positive impact compare models and finite element model of the impulse response.
[RESULTS,CHAIN,S2CHAIN,SSCHAIN] = MCMCRUN(MODEL,DATA,PARAMS,OPTIONS)
sum-of-squares function 'model.ssfun' is called as
ss = ssfun(par,data) or
ss = ssfun(par,data,local)
instead of ssfun, you can use model.modelfun as
ymodel = modelfun(data{ibatch},theta_local)
The impact of historical value used in the study
MCMC sample parameters when the impact influence on the bay, measurement noise level of 5%.
out = mcmcpred(results,chain,s2chain,data,modelfun,nsample,varargin)
parind = results.parind;
local = results.local;
theta = results.theta;
nsimu = size(chain,1);
nbatch = results.nbatch;
%MCMCPLOT Plot mcmc chain
mcmcplot(chain ,1:4,[],'pairs')
Histogram and the determined parameter fit marginal PDF, the measured noise level of 5%.
FIG identified normal parameters, measurement noise level of 5%.
Bay identification impact history, 90% confidence interval, measured noise level of 5%.
Determining the impact force history, 90% confidence interval, measured noise level of 5%.
Determining the impact force on the ribs history, 90% confidence interval, measured noise level of 5%.
Impact position in response to identified, 90% confidence interval, measured noise level of 5%.
Comparative impact energy measurement and identification.
in conclusion
This study proposes a Bayesian statistical methods for impact location and impact of historical recognition of reinforced composite board. By using the first set of parameters represents a shock load to convert the parameter identification problem issues. In the identification process characterized by comprising a reinforcing plate composite dynamic response forward models known impact shock loads. By combining the measurement data and a priori information, Bayes' theorem is used to update the probability distribution of the parameters. In particular, MCMC methods for sampling the posterior distribution parameters to estimate the impact.
Have questions please contact us!
Big Data tribe - Chinese professional third-party data service providers to provide customized one-stop data mining and statistical analysis consultancy services
Statistical analysis and data mining consulting services: y0.cn/teradat (Consulting Services, please contact the official website customer service )
QQ：3025393450
QQ exchange group: 186 388 004
[Service] Scene
Research; the company outsourcing; online and offline one training; data reptile collection; academic research; report writing; market research.
[Tribe] big data to provide customized one-stop data mining and statistical analysis consultancy
Welcome to elective our R language data analysis will be mining will know the course!
