Modelling, verification and validation

Text Box: Verification, validation and quality management One day short course Introduction The modelling, verification, validation and quality management activities are perhaps the most crucial aspects of the simulation and analysis process. A model yielding erroneous results may engender serious consequences, not least in terms of product safety, legal liability and additional cost. This course identifies the common sources of error, uncertainties and mistakes that inevitably arise in the computational modelling process, and highlights methods for avoiding or treating them using best-practice modelling, verification, validation and quality management techniques. The course begins by describing a typical modelling process by abstracting an idealised description of a real-world object and its characteristics in the form of a mathematical model. While the formal mathematical modelling stage is often seemingly bypassed as the computational model is constructed directly in software, the underlying idealisation assumptions nevertheless directly affect the validity of the model, and need to be clearly identified. A structured, best-practice approach to modelling which enables systematic model verification, validation and management, is presented. Model input data needs to be of sufficient accuracy to support the required solution accuracy. Methods which can be used to estimate the required accuracy of the data are discussed. Model verification, which is defined as confirmation that the computational model conforms to the specified mathematical model and that its solution is computationally accurate, is discussed along with model validation, which is defined as ensuring that the idealisation assumptions and data are correct, thereby providing a basis from which to assess the fitness of the model for its intended use. The course proceeds to explain how the computational modelling process, including verification and validation activities, can be systematically managed using the quality management system ISO 9001 and NAFEMS QSS 001 (a supplement to ISO 9001 for engineering simulation) to minimise the opportunity for procedural error. The quality management content of the course can be used by attendees to assist in the development and implementation of a quality management system for simulation applications, or in adapting to or integrating with, a wider quality system. Emphasis is placed on understanding the basic concepts of the ISO 9001 and QSS 001 process-based quality models and their practical application to analysis management. Formal quality management system requirements are discussed along with issues relating to the design, implementation, maintenance and, if required, formal certification of the system. Finally, the effect of the implemented quality system on simulation product quality, and on the working practices and culture of the organisation, will be discussed. A set of rudimentary quality management procedures for engineering simulation will be supplied. Course content 1 Engineering simulation 1.1 Mathematical modelling 1.2 Numerical approximation 1.3 Scientific computing 1.4 Modelling cycle 2 Sources of error 2.1 Procedural error 2.2 Modelling error 2.3 Computation error 3 Control of computation error (model verification) 3.1 Principles of model verification 3.2 Software error 3.3 Numerical error 3.4 Convergence of results 4 Control of modelling error (model validation) 4.1 Principles of model validation 4.2 Idealisation error 4.3 Data error 4.4 Consistency of results 4.5 Structured modelling 5 Control of procedural error (quality management) 5.1 Principles of quality management 5.2 Quality management system 5.3 Management responsibility 5.4 Resource management 5.5 Product realisation 5.6 Measurement, analysis and improvement 5.7 Quality documentation and records 5.8 Quality certification 5.9 Degree of verification, validation and quality control Appendix A : Quality management procedures for engineering simulation

Presenter: Jonathan Smith

Duration: One day

Synopsis

Computational models need to be shown to be fit for their intended use. Moreover, sound and systematic management of the computational modelling process is essential if the opportunities for errors and mistakes are to be minimised.

This course identifies the common sources of error and uncertainty that inevitably arise in the computational modelling process, and discusses systematic methods for controlling those errors and uncertainties though model verification, validation and quality management techniques.