Virtual modeling with software verifies components concurrent with other development tasks, while virtual integration can prevent inferior quality or missed deadlines caused by detecting errors too late in the game — during physical testing.

In short, virtual integration allows engineers to find and fix errors earlier, when they are easier and cheaper to fix. Modeling interactions reveals ambiguous or contradictory requirements. In addition, when individual component models are included in full-system models, it is easier to assess how component iterations affect overall behavior.

Three requirements

Let us consider embedded systems — good candidates for virtual integration. Embedded systems boost functionality for many designs, but at the price of increasing complexity. Normally, to confirm that an embedded system meets quality and design requirements, engineers must conduct time-consuming and resource-intensive design and code reviews, plus build complex test systems. Often, this verification represents more than 50% of total development cost.

To bring costs down, automotive, aerospace, and industrial engineers increasingly use virtual integration through model-based design, a process in which software improves designs and automatically generates code to streamline implementation. No actual system hardware is used; a program organizes and executes testing.

Virtual verification with model-based design has three parts.

1. An executable system specification is a full-system model or set of models that can be probed and investigated, usually through simulation. It includes models of relevant subsystems (control algorithms, mechanical and electrical components, hydraulics, and so on) detailed enough to reflect functional behavior. A system model should also include operating environment or plant models with actuator and sensor models for control applications, or transmission environment models for communications.

2. Requirements-based tests exercise executable system specifications to validate requirements. Tests are based on functional requirements that can be modeled as simulation input signals or sequences and, ideally, expected outputs. Some software can even show which parts of a model are actually exercised during a sequence of simulations.

3. The final element for virtual integration is a test and simulation executive — an environment for managing all tests, results, and running simulations. This environment is also used to author tests, compare those to models, and generate results for analysis. (Here, aerospace and automotive engineers often leverage computing clusters — multiple distributed computers — to complete many complex simulations in a reasonable timeframe.)

Typical versus model design

A development process using model-based design automatically generates code from detailed executable-specification models — so systems and their individual parts can be verified. First, written requirements are used to create the system-level specification, which in turn drives subsystem designs. Next, the subsystems are tested and integrated. Finally, the complete embedded system is tested; errors are detected as they are introduced.

Building the foundation

Model-based design: In traditional development, fundamental questions go largely unanswered until actual physical components and systems are tested. Virtual integration and testing creates a second V that features multiple feedback loops — shown by the green arrows — is more effi cient and accurate.
Select figure to enlarge.

Most organizations utilize modeling, but few do it systematically, so the full benefits of simulation aren't realized. Where engineers buck this trend and insist on testable models, complete system models verify that requirements make sense, and that designs work.

In short: Plant parameters, inputs, and environmental factors vary, and time and budget constraints often limit how much variability can be incorporated into tests on actual hardware. On the other hand, centralized simulation accelerates individual tests and automates multiple tests across a range of input parameter values. In addition, limits and potentially destructive tests can be safely investigated virtually. Exploring all parameters in simulation also minimizes the number of critical tests that must be run in real time.