The objective of the optimize phase is to generate a detailed product design. The starting point is the high level design, the architecture, developed in the design phase. In the design phase the specifications of the system requirements were translated to specifications for the building blocks. This should have been done in such a way that it resulted in a design with no capability gaps, i.e.:

(i) the specifications for the building blocks must be realistic and achievable and

(ii) the specifications of the building blocks must ‘add-up’ to the specifications of the end-product

**Detailed design**

Based on the architecture, the building blocks are further designed in detail. This means that the components for the building blocks are specified, the sub-components of the components are specified and that the processes producing the sub-components are described.

**CTQ flow down**

A deliverable from the Identify phase is an overview of all system requirements (CTQ’s) derived from the target customer needs. An overview of all the parameters in the design underlying these system requirements at the different design levels, such as end product-, building block–, component–, sub component–, and process-level, and their relations can be summarized in the so-called CTQ-flow down. These relations can be described qualitatively by ‘very high impact’, ‘high impact’, ‘moderate impact’, ‘weak impact’ or quantitatively.

The quantitative approach is again described as a set of transfer functions. This means that the relations between all critical parameters throughout the design are linked to each other with transfer functions. Within DfSS such a set of transfer functions is seen as the highest form of product technology knowledge.

**Integral optimization and robust design**

For those parts of the design where the relations between the design parameters are described by transfer functions, the technique of mathematical optimization can be used to make:

- Integral optimum designs.

The set of transfer functions usually contain many parameters, having complex non-linear relations and several constraints. This makes it difficult, even with expert physics knowledge, to set the best design. It is our experience that mathematical optimization gives much better designs than the engineers can achieve using their expert knowledge. - Robust designs.

Robustness means that the design is as less sensitive to unwanted variations as possible. - Make trade-offs.

Frequently designs contain conflicting parameters, that is, the best settings for one parameter are not good for another parameter. Mathematical trade-offs help to solve this problem. The case study for robust design and tolerance analysis (stiffness and torque) gives an example of such a trade-off.

**Capability prediction and Yield checks**

In case transfer functions are made for each building block, they can be connected to make an integral mathematical model for the total design. As a result this can be used to make capability predictions in the same way as is done in the design phase for the high level design. Of course, we should not only make predictions for short term production spread, but also mid- and long term production spreads must be taken into account in the capability predictions.

In addition to a capability prediction based on mathematical models, also small series of prototypes are build on a lab-line or a pilot line. The measurements of the first series of products give the first capability and yield insights on real products.

**Reliability**

As stated before, a study of the life time, i.e. reliability, of the product is often postponed to later development phases when several products are available. However, this is frequently too late or too costly to make design changes. Starting earlier with reliability may prevent this. As soon as the first products are available, one can study the dominant failure mechanisms by doing some Highly Accelerated Life Tests (HALT). HALT give the dominant failure mechanisms under stress but do not predict the life time. To this end, studies have to be started that determine the impact of these failure mechanisms on life-time, i.e. life-time models.

**Risks**

Also in this phase the risks need to be assessed. The product FMEA created in the design phase needs to be updated based on the work done and knowledge obtained during the optimize phase. For example, a high risk on a CTQ very sensitive to variation has been addressed by the optimized, robust design. In this phase also the process risks need to be assessed using an FMEA, given amongst others attention to the transition from prototyping to mass production.

Last but not least, also the project related risks need to be re-evaluated.

**CQM**

Building transfer functions and connecting them for complex systems is usually beyond de limits of Excel. Moreover finding the optimum and robust design settings cannot be done manually and also ad-hoc algorithms will not lead to the best design. CQM developed the Compact-CO module for this task. This module has proven to come up with optimized designs that beat any other design.

Predicting and modeling life-time of products is also a core competence of CQM

Statistical analysis of the first prototype series can be done in Excel using the add_in CQM_EfP, see more at: /cqm_efp.