Model Order Reduction for Wave Propagation Soda Hall波传播汽水大厅的模型降阶.pptVIP

下载本文档

1
0
约1.24万字
约 29页
2017-03-08 发布于上海
举报
版权申诉

Model Order Reduction for Wave Propagation Soda Hall波传播汽水大厅的模型降阶.ppt

1、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
3、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
4、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
5、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
6、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
7、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

Model Order Reduction for Wave Propagation Soda Hall波传播汽水大厅的模型降阶

Midway Shot Full waveform inversion In this application we have seismic data collected (say) at the free surface on a number of receivers: Sio(tj) and a parameterized velocity model v(a). Starting from an initial guess v(a0) we use the HF and MOR codes to generate the calculated response at the receivers and use a least squares data fit to improve upon the values of the velocity parameters. For this purpose we employ a Levenberg-Marquardt code that approximates gradients with finite differences (LMDIF from MINPAK). Regularization/Convexification/ContinuationRCC For difficult inverse problems is recommended to use regularization, i.e., add to the misfit functional a term of the form: mina (1-l)||So – Sc(a)||22 + l ||a – a0||22, that penalizes large changes in the unknown parameters from some reference value. We observe that the penalty term is a convex function. Also, this is the form for homotopy continuation, where we start from a known solution (a0, for l =1) and move l incrementally, towards 0, the problem we really want to solve. Multimodal Response Function We consider a simple 3 constant layers model and vary only the velocity in the first layer. We plot the response function for a range of values of v1 and of l. Some Preliminary Inversion Results We consider the same 2D, 3 homogeneous layer problem. Model size: 3500x3500. The data is generated by the HF code and corresponds to a shot at x=1600. There are 200 receivers spaced by 10m, starting at x=700 (400 samples per receiver). l was set to 0.05. The purpose is to recover the (constant) velocity in the layers, when starting away from the values that generated the synthetic data. This results in a 80000 x 3 nonlinear least squares problem. MOR uses 108 snapshots from the first HF simulation at a0. Numerical results Target Initial Final HF Final MOR-2 v1 1500 1400 1500 1500 v2 2500 2600 2500 2500 v3 3500 3425 3500 3500 Evals. 67 37 Total time 5025” 397.75” Final residual 0.0 0.0 Speed up = 5