Abstract: Under the current regulation system, the statistical model based on attribute sampling is a common method for the quality assurance of earth-rockfill dam compaction; its key indicator is the specific threshold of density defect as is usually recommended. However, a statistical model is inevitably faced with two statistical errors－a false-negative problem and a false-positive problem. A careful examination of the optimization method of sampling schemes is necessary for strict quality assurance. This paper develops two improved models based on different risk criteria－featuring in consumers first, and risk balancing of both parties for the risk appetite of contractor and consumer respectively. And a Monte Carlo based simulator is provided, which can achieve significant improvement on attribute sampling by combining with a space optimization algorithm. Application to an engineering case verifies the practical effect of our optimization method. This is a promising method for ensuring the rights and interests of both parties, through reducing the two statistical errors and limiting the impact from the spatial correlation of samples.

Key words: construction quality control, earth-rockfill dam, compaction quality, attribute sampling, sample size, optimization method