The Verify phase focuses on the preparation for mass production and realizing the market introduction. In addition to designing the product, also the production processes have to be developed. Provided with a feasible product design from the optimize phase, we now need a process capable for mass production. Proceeding from feasibility to capability is the objective of the Verify phase.
Splitting up the R&D process according to DfSS into the DIDOVM phases suggests that this is a sequential approach. Reality shows that the opposite is true given the fast, competitive ways of doing business. So development of the production processes usually starts in parallel to the development of the product design. In particular when the production processes use new technologies.
Ramp up, baselines and Design of Experiments
The first product prototypes are usually built on a laboratory production line. As the R&D-project progresses, production of prototypes and engineering samples shifts to a pilot line, next to a semi-industrial line and finally to the industrial line. Meanwhile, the engineering-production volume increases all the time. In the end , the products have to meet the specifications with high capability, e.g Cpk as metric, and with processes having high yields, e.g. rolled throughput yield as metric.
The objective of DfSS is predictable and capable processes. This will be achieved by in depth and fact-based understanding of the critical parameters that will have the biggest influence during production. Insufficient testing is a frequent cause for unexpected technical problems during production or at the customer. Therefore, the DfSS credo is testing, testing and testing. Alternating “Baselines” and “Design of Experiments (DoE)” will do the job:
- Baselines are production runs under frozen process conditions. These frozen baseline settings are the best proven settings so far. It is only allowed to switch to new baseline settings when it is statistically proven that the new settings perform better.
- Better settings are found by performing sound experiments. To this end the theory of Design of Experiments is needed. Experiments which are set-up according to DoE, pinpoint the sensitive parts and sensitive areas in designs that cause the gaps in performance and yield.
Measuring and testing
Testing means that we need test plans that describe what should be measured, how frequently and by which method. Within DfSS this is according to the V-model in combination with statistical concepts.
The V-model states that for all parameters in the CTQ-flow down and at all levels, so system-, building block-, component-, sub component-, and process parameter level, there must be a test plan. Statistical techniques will tell how many and how many times products and components have to be sampled in order to give the required accuracy. The measurement methods of all these parameters should be evaluated with the so called Measurement System Evaluation technique.
At the end of the Verify phase the capability for mass production needs to have been validated. To achieve this, the DfSS approach is based on fact-based evidence throughout the project. But how should this evidence look like? The CQM approach to DfSS gives answers.
- Evidence of supplier quality.
In many industries, the traditional AQL-system is still used. Most people still don’t know that the AQL system protects the supplier and not the customer. (Related post)
For this reason CQM developed a system that really controls the incoming supplier quality.
- Evidence at milestones.
Every R&D department has a milestone procedure. And in every milestone procedure there is a chapter on product and process quality. Nevertheless, there are very often discussions at milestone meetings whether the quality goals are met. Basically, this is because the criteria are not statistically sound or not uniquely defined. Therefore CQM described milestone criteria along the lines of the DfSS approach.
- Evidence that the processes will keep their performance.
A good performance during the engineering stages does not imply that the process will forever perform as good. In generally, processes will not perform constant when no precautions are taken. In industrial processes, Statistical Process Control (SPC) is commonly applied to safeguard processes against drifts and trends. Typically, SPC should be implemented during the Verify phase.
During the Verify phase, the reliability can be tested on the final product design and also on products coming from the final industrial line. In general the following reliability tests will be performed:
- Stress (HALT or MEOST) tests are used to compare about 10 products with 10 reference products , either from own company or from a competitor . The objective is to identify the weakest components.
- Stress (ALT, or accelerated use) test and real time tests are performed to predict the lifetime of the products under use conditions.
Based on the insights from this phase the product FMEA is updated and action planning is addressed.
During the Verify phase the first production series generate data and hence the classical statistical techniques can be applied to (i) develop sampling plans, (ii) extrapolate performance to the industrial phase, (iii) draw sound conclusions on cause and effect and (iv) discover trends and deviations in data full of noise. All of this is CQM’s core business.