Evaluation of Site Investigation Sufficiency by Bayesian Compressive Sampling

Description

Unexpected site condition presents a major risk in geotechnical construction, and itoccurs when the expected ground condition interpreted from site investigation issignificantly and adversely different from the actual ground condition found duringconstruction. It often causes significant project delays and expensive over-budget. TheGuangzhou-Shenzhen-Hong Kong Express Rail Link construction project in 2014, forexample, had a budget of HK$67 billion, but suffered from at least a two-year projectdelay and required an extra HK$20 billion. Unexpected site condition was listed as amain reason for the delay and overspending. The problem of unexpected site condition isobviously due to insufficient site investigation. To avoid this problem, the sufficiency ofsite investigation should be evaluated, and proper sufficiency must be achieved. However,NO scientific or quantitative method currently exists to evaluate the sufficiency of siteinvestigation or to quantify statistical uncertainty that occurs when interpreting groundcondition from sparse and limited site investigation data. Conventional statisticalmethods, or even geostatistical methods, only have limited applications in geotechnicalsite investigation, as geotechnical data are spatially correlated, but only sparselymeasured during site investigation.This project aims to develop novel statistical methods and tools (e.g., statistical chartsand computer software) to quantify statistical uncertainty associated with interpretationof spatially correlated, but sparsely measured, site investigation data and to evaluate thesufficiency of site investigation. The proposed methods are based on an innovativesampling theorem called compressed/compressive sensing/sampling (CS), which wasawarded the prestigious Shaw Prize (widely regarded as the “Nobel of the East”) inmathematical sciences in 2013. The PI has performed preliminary studies on CS-basedinterpretation of site investigation data and obtained promising results. Major researchcomponents of the project include: (1) theoretical development of Bayesian CS (BCS)methods for statistical uncertainty quantification, (2) development of BCS-basedmethods and statistical charts for evaluating site investigation sufficiency, (3) calibrationof the developed methods and statistical charts using field data and observations, and (4)development of computer software for the proposed methods.For the first time ever in geotechnical engineering, this project will provide rational andquantitative methods for evaluating site investigation sufficiency. It will also offerstatistical charts and computer software to facilitate application of the proposed methodsin engineering practices. The research outcome will be beneficial to geotechnical practicein Hong Kong, where many mega-infrastructure projects are currently in progress, andthroughout the world.?