Estimating probability curves of rock variables using orthogonal polynomials and sample moments

Abstract

A new algorithm using orthogonal polynomials and sample moments was presented for estimating probability curves directly from experimental or field data of rock variables. The moments estimated directly from a sample of observed values of a random variable could be conventional moments (moments about the origin or central moments) and probability-weighted moments (PWMs). Probability curves derived from orthogonal polynomials and conventional moments are probability density functions (PDF), and probability curves derived from orthogonal polynomials and PWMs are inverse cumulative density functions (CDF) of random variables. The proposed approach is verified by two most commonly-used theoretical standard distributions; normal and exponential distribution. Examples from observed data of uniaxial compressive strength of a rock and concrete strength data are presented for illustrative purposes. The results show that probability curves of rock variable can be accurately derived from orthogonal polynomials and sample moments. Orthogonal polynomials and PWMs enable more secure inferences to be made from relatively small samples about an underlying probability curve.

Keywords

rock variables / probability curve / orthogonal polynomials / conventional moments / probability-weighted moments

References

[1] DengJian, YueZhong-qi. Pillar design by combining finite elements, neural networks and reliability: a case study of Feng Huangshan Mine in China[J]. International Journal of Rock Mechanics and Mining Sciences, 2003, 40(4): 585-599

[2] QuekS T, LeungC F. Reliability-based stability analysis of rock excavations [J]. International Journal of Rock Mechanics and Mining Sciences, 1995, 32(6): 617-620

[3] MelchersR EStructural Reliability Analysis and Predictions (Second edition)[M], 1999, New York, John Wiley & Sons Inc

[4] ZhuYu-xueSlope Reliability Analysis[M], 1993, Beijing, Metallurgy industry Press(in Chinese)

[5] PearsonE S, JohnsonN L, BurrI W. Comparison of the percentage points of distributions with the same first four moments, chosen from eight different systems of frequency curves [J]. Commun Statistics B, 1979, 8: 191-195

[6] KennedyC A, LennoxW C. Solution to the practical problem of moments using non-classical orthogonal polynomials, with applications for probabilistic analysis[J]. Probabilistic Engineering Mechanics, 2000, 15(4): 371-379

[7] DengJian. A distribution-free method using maximum entropy and moments for estimating probability curves of rock variable[J]. International Journal of Rock Mechanics and Mining Sciences, 2004, 41(3): 376-376

[8] PandeyM D. Direct estimation of quantile functions using the maximum entropy principle [J]. Structural Safety, 2000, 22(1): 61-79

[9] SiddalJ NProbabilistic engineering design principles and applications[M], 1983, New York, Marcel decker Inc

[10] BadinelliR D. Approximating probability density functions and their convolutions using orthogonal polynomials [J]. European Journal of Operational Research, 1996, 95(1): 211-230

[11] GreenwoodJ A, LandwehrJ M, MatalasN C. Probability-weighted moments: definition and relation to parameters of several distributions expressable in inverse form [J]. Water Resources Research, 1979, 15(5): 1049-1054

[12] HoskingJ R M. L-moments: Analysis and estimation of distributions using linear combinations of order statistics[J]. Journal of the Royal Statistical Society, Series B, 1990, 52(1): 105-124

[13] JaegerL G, BakhtB. Lower fractiles of material strength from limited test data [J]. Structural Safety, 1990, 12(7): 67-75