Introduction

This manual will attempt to motivate the use of an automated build system for the purposes of computational science and engineering. As part of this motivation, the surrounding computational practices of version control, documentation, compute environment management, regression testing, and data archival will also be addressed as applied to the practice of computational engineering. Specifically, this manual intends to motivate the adoption of these traditional software engineering practices for use in research and production engineering simulation projects.

This manual is not the first such effort in the greater scientific computing community. In fact, the authors relied heavily on the lesson plans of the Software Carpentry, established to teach computing skills to researchers in 1998 [1]. As the intention for this manual is to motivate the fundamental practices of engineering computing, it will not attempt to fully teach the underlying concepts and will instead reference the well designed lesson plans of the Software Carpentry.

This manual also does not intend to replace the modeling and simulations standards documents in fields related to computational engineering. The American Society of Mechanical Engineers (ASME) standards for computational solid mechanics [2] and computational fluid dynamics [3] and the National Aeronautics and Space Administration (NASA) standard for models and simulations [4] with associated handbook [5] are excellent resources for practicing computational engineers. Where topics in such standards overlap with the discussion in this manual, this manual should be taken as one possible solution to a subset of the standards’ requirements but not as a replacement text. Readers are encouraged to seek out the standard adopted by their organization or professional society for adoption in their work. Regardless of the chosen standard of your field, the NASA standard is recommended by the authors for the discussion about model documentation, development, testing, verification, and release requirements.

Where possible, this manual will explain to general computing practices and concepts and limit discussion of specific software implementations to examples or vehicles for practice in concrete application. The specific software taught by the Software Carpentry curriculum is an excellent starting point to learn the core concepts of computational engineering. However, the authors have found that applications to engineering simulation and analysis require translation of these software development concepts into the language and workflows of computational engineers. Adopting these computational tools may require engineers to re-imagine their workflows in some combination of traditional engineering and software concepts.

It has also been necessary to extend existing software build systems for engineering practices beyond the simple wrapping of engineering software execution. Where necessary, examples of specific software and their method of extension to engineering simulations will be given, with reference to the User Manual for recommended practical use. Where this manual relies on specific implementation examples, it should be understood that the practicing engineer may find that different software is more amenable to their specific work. The overall collection of computational practices is more important than any specific software implementation. The ability to recognize which concepts are implemented by a software package will make a practicing engineer agile to changing project needs, computing resources, numeric solvers, programming languages, and even available funding.

Core Concepts

The core concepts of computational engineering practice are listed below in the rough order of project maturity. That is, those things listed near the beginning of the list should be implemented early in the life-cycle of all projects, regardless of expected project growth, re-use, or impact. Particularly, version control and documentation should be implemented at the very beginning of a computational project, even with the earliest project brainstorming session. As the foundational practice of computational engineering, version control fills the role of a laboratory notebook and the strict practice of detailed and regular entries are equally important.

Those practices at the end of the list are more costly to implement and therefore more valuable for mature or long-lived projects. However, an engineer familiar with the computational tools implementing these practices will find that even small, short-term projects benefit from implementing all of these concepts. In production engineering environments, it is beneficial to implement the full range of computational engineering concepts in a “stub” repository as a starting template for every project. For this purpose, WAVES provides ModSim Templates following the tutorials as a starting point, but it is expected that users will want to tailor a template specific to their application.