Applications to the Design of Trustworthy Software

Having reviewed the elements of Taguchi Methods as they relate to software design, we now want to pull everything together as a set of guidelines for the use of these Methods in the earliest stages of design. Chapters 18 and 19 of Design for Trustworthy Software by Jayaswal and Patton, published by Prentice Hall, show how Taguchi Methods can be adapted for validation, verification, testing, evaluation, integration, extension, and maintenance to ensure trustworthy software. For software development, design is the primary activity, but we include requirements development, functional specification, and technical design as the three phases of the design process. Clearly Taguchi Methods apply only after the first round of these three phases and after a design is proposed that satisfies the user’s requirements, meets the specification, and now needs to be “bulletproofed,” as the programmer usually puts it. Every Taguchi optimization is different because every product’s performance target is different. Thus, the design engineer and his or her team need to pursue the steps shown in Table 1. They are the best equipped to identify software performance characteristics and noise factors, determine their levels, and, with the help of statistics, analyze the performance statistics and validate the experiment(s) with a follow-up run to confirm the predications.