With more and more attention on degradation process, we need the degradation model to be accurate
all over the time rather than only at some specific moments. However, the traditional validation metric only
estimates difference of static features. A validation method proposed in this paper uses hypothesis testing to
identify whether the distributions of experimental measurements and simulation results are consistent. Then, based
on the deviation between sample means, a global validation metric which reflects the difference of degradation
process between computational model and physical system all over the service time is derived from the statistics
of deviation between sample means. Furthermore, curve fit method for discrete experimental measurements is
introduced. The case of electro-hydraulic servo valve is studied, and the results show that the proposed validation
metric is appropriate for the validation of degradation model with dynamic performance output.
YANG Chun-boa (阳纯波), ZENG Sheng-kuia,b (曾声奎), GUO Jian-bina,b* (郭健彬)
. Validation Metric of Degradation Model with Dynamic Performance[J]. Journal of Shanghai Jiaotong University(Science), 2015
, 20(3)
: 302
-306
.
DOI: 10.1007/s12204-015-1626-7
[1] Balci O. Verification, validation, and certification of modeling and simulation applications [C]// Proceedings of the 35th Conference on Winter Simulation.New Jersey, USA: [s.n.], 2003: 150-158.
[2] Sargent R G. Verification and validation of simulation models [C]// Proceedings of the 37th Conference on Winter Simulation. New Jersey, USA: [s.n.], 2005:130-143.
[3] Oberkampf W L, Barone M F. Measures of agreement between computation and experiment: Validation metrics [J]. Journal of Computational Physics,2006, 217(1): 5-36.
[4] Rebba R, Mahadevan S, Huang S. Validation and error estimation of computational models [J]. Reliability Engineering and System Safety, 2006, 91(10):1390-1397.
[5] Bachoc F, Bois G, Garnier J, et al. Calibration and improved prediction of computer models by universal Kriging [J]. Nuclear Science and Engineering, 2014,176(1): 81-97.
[6] Kolluri R. Provably good moving least squares [J].ACM Transactions on Algorithms, 2008, 4(2): 1-25.
