On System Identification for Accelerated Destructive Degradation Testing of Nonlinear Dynamic Systems
Accelerated destructive degradation testing is considered with the objective of reproducing high fatigue incidents for a severely nonlinear system in a lab environment. In the lab, a test specimen is mounted on servo hydraulic actuators which are then use
- PDF / 4,237,193 Bytes
- 30 Pages / 439.36 x 666.15 pts Page_size
- 47 Downloads / 206 Views
On System Identification for Accelerated Destructive Degradation Testing of Nonlinear Dynamic Systems Jacq Crous, Daniel Nicolas Wilke, Schalk Kok, Ding-Geng (Din) Chen, and Stephan Heyns
Abstract Accelerated destructive degradation testing is considered with the objective of reproducing high fatigue incidents for a severely nonlinear system in a lab environment. In the lab, a test specimen is mounted on servo hydraulic actuators which are then used to induce the same response in the system as was measured in field tests. Finding the inputs to the actuators that accurately induce the measured response in the system is crucial to the integrity of the testing procedure. The problem is an inverse problem, and often exhibits ill-posed characteristics. To this end a new method for system identification from time series data is developed and is shown to outperform current methods such as different variants of NARX and Hammerstein-Wiener models. From the results obtained it is concluded that an alternative method of data generation for accelerated destructive degradation on severely nonlinear systems in a lab context is required. Three methods are developed and tested on simulated data and it is shown that a prototype bootstrapping strategy is superior: using 400,000 data points generated by this strategy the input signals were predicted with mean square errors of 5.08e-4. Keywords Accelerated destructive degradation • Time series analysis • Spanning basis transformation regression • Nonlinear system identification
J. Crous () • D.N. Wilke • S. Kok • S. Heyns Department of Mechanical and Aeronautical Engineering, University of Pretoria, Pretoria, South Africa e-mail: [email protected]; [email protected]; [email protected]; [email protected] D.-G. (Din) Chen School of Social Work & Department of Biostatistics, University of North Carolina, 27599, Chapel Hill, NC, USA Department of Statistics, Univeristy of Pretoria, Pretoria, South Africa e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2017 D.-G. (Din) Chen et al. (eds.), Statistical Modeling for Degradation Data, ICSA Book Series in Statistics, DOI 10.1007/978-981-10-5194-4_17
335
336
J. Crous et al.
17.1 Introduction Accelerated testing is ubiquitous in engineering manufacturing and certification to rapidly obtain reliability information. Numerous approaches to achieve this have been developed depending on the application under consideration. These include use-rate acceleration, aging-rate acceleration and increased intensity acceleration [14]. Use-rate acceleration is particularly applicable to systems that are either not continuously used or in continuous use but only exposed to discrete events that significantly affect the life of the system. To accelerate this scenario a lab scale test that can be used to expose a system more frequently to these events is often required. This approach is valid as long as the time-scale and cyclic rate through these events do not affect the cycles-to-failure distribution. Although use-rate accele
Data Loading...
