Automatic Validation of the Offshore Operators Committee Discharge Model and Application to Predicting Drilling Solids Accumulation on the Sea Floor

摘要

Discharge models must be validated by comparing their predictions with experimental data. Such comparisons improve user confidence in model predictions. The validation process involves significant work for each case tested and requires tedious labor. This paper describes an automated validation system and its use in validating the offshore operators committee (OOC) Mud and Produced Water Discharge Model. Once the validation system is set up, very little additional work is needed for repeated validations to test the model after changes related to maintenance and development. The validation system provides a complete record of all validation methods, data, and results. The principle benefits of the automated validation system are: the combined validation tests are completely documented with an HTML report, the tests are easily repeated, the system quickly reveals flaws arising from model maintenance and development activities, the system can be adapted to other numerical models containing standalone executable modules that read and write text files. The automatic validation system consists of several parts: (1) a hierarchical arrangement of data to segregate individual experiments in separate file system directories; (2) command scripts to run validation tests in each directory (model runs, statistical comparisons of predictions and observations, plots of predictions compared with observations); (3) a top-level script to summarize overall comparison statistics and scatter plots; (4) a report generator to assemble validation results in a linked set of HTML pages with plots; and (5) a tool to compare validations run at different times (e.g., to compare predictions of different versions of the model). Experiments included in the validation system are summarized briefly. One laboratory and one field experiment concerned with particle deposition on the sea floor were added during the development of the validation system. These cases are described in more detail.