Modelling of the behavior of marine oil spills: applications based on random walk techniques

A numerical model has been developed to simulate the transport and fate of oil spilled at sea. The model combines the transport and fate processes of spilled oil with the random walk technique. Oil movement under the influence of tidal currents, wind driven currents, and turbulent eddies is simulated by the PLUME RW dispersion model developed by HR Wallingford. The weathering processes in the model represent physical and chemical changes of soil slicks with time, and comprise mechanical spreading, dispersion, evaporation and emulsification. Shoreline stranding is determined approximately using a capacity method for different shoreline types. This paper presents details of the model, and describe the results of various sensitivity tests. The model is suitable for oil spill contingency planning.

基于EFDC模型的感潮江段溢油事故风险预测

为了在溢油事故发生后能够立即采取有效措施控制和减轻油污染,降低事故危害,采用环境流体动力学模型(EFDC)对长江下游靖江段某码头进行溢油事故风险影响预测。通过EFDC准确地模拟出该江段的二维流场,在流场、风场以及复杂地形等综合条件下,采用拉格朗日质点追踪法计算油品入江后油膜漂移轨迹以及到达、离开下游保护区的时间。结果表明:油膜沿水流方向逐渐被拉伸,覆盖面积逐渐增大,且受长江地形条件影响;在感潮江段,潮流场对油膜漂移行为的影响占主导作用,同时风场也会影响油膜的漂移行为。受涨落潮的影响,油膜向下游来回震荡漂移;溢油事故发生时刻的流场不同,油膜往下游漂移的速度也不同,当溢油事故发生在落潮时,油膜往下游漂移速度更快;不利风向时风速越大,油膜往下游漂移的速度越快。

溢出代码和访存压力敏感的快速机器学习

基于遗传算法提出了溢出代码和访存压力敏感的机器学习来调试寄存器分配的权值函数。不同于以往采用目标程序的运行时间作为适应值,通过静态分析寄存器分配产生的溢出代码和基本块中的访存压力来构建适应值,以减少学习时间。这些分析被限定在热点函数中,在保证适应值精度的同时进一步加快了学习速度。实验表明,快速学习仅需要考虑热点函数的编译时间,整个CPU2000CINT测试集在5 h内即可学习完毕。大部分CPU2000CINT测试例子的性能得到了提高。其中perlbmk的性能提升最高可达到7.2%。