I think, at least, the correct model should be tested fully. In addition to "validate", should also meet what condition? I can think of is "verifiable", "verifiable" is the premise of fully validated.
Model will specially introduce "verifiable", this article first tell me about the model code generation which validation can be done before?
According to the model is run to differentiate, can be divided into static and dynamic authentication to verify two kinds big.
Static verification assessment, modeling standard examination, formal verification, etc.
1 model review
Evaluation is the quality system requirements, usually after our model picture, need to review the model by means of review whether to implement the corresponding requirements. Then, review the moment should be how to grasp? Whether the finished model followed by review went to do? Personally, I think, if you have tools can help us to realize static checking, why don't we do static checking first? Do the benefits of static checking is to put the first can be found through the tools of software defects eliminate before the review, so that you can avoid the review of waste time on such problems.
2 modeling standards inspection
Now through MATLAB SimulinkVerification&Validation provided ModelAdvisor (hereinafter referred to as SLVnV), can realize the static modeling standards inspection, it is also easy to implement, and tools also can be customized function, can be in a lot of check items in the selected SLVnV provide check set suitable for their own development team, there may be some check is ModelAdvisor did not provide, in this case, also can through the way of MATLAB to write the custom check items.
3 formal verification
MATLAB provides SimulinkDesignVerifier (hereinafter referred to as the SLDV) product, the model can be formalized validation. SLDV can check in the model for integer overflow, or whether there is logic. These two types of errors are very easy to review and functional tests miss mistake, in fact, once found that the two types of mistakes in the software, repetition or locate the two types of errors are very difficult, and SLDV can or possible error by analyzing the model of the link. In addition, it is important to note that in the process of model evaluation or functional test, we can find some software defects, and found these defects, we need to modify the model, the modified model could be introduced again data errors such as overflow or dead logic, suggest to do functional testing, so use SLDV again check the model.