Abstract
Use of fractional experimental design gives enormous save of cost in research and development. The price of such savings is confounding of factors.
Two 2(7-3) fractional experiments have been done, one of it in mechanical production (hardening of projectils) and the other with a biological issue. Conditions for biomass production of simple plants was been investigated. Results have been analized with classical statistic methods at all experiments. Planning, executing, documentation of the data and teaching of research personnal has been done and discussed according to routines suggested by the NS-ISO 9001 (EN 29 001) norm. Follow up runs (with 23 full factorial designs) has been designed based on significances found in the first experiments.
Response surface plots has been created, but only verified for the biological issue.
The industrial experiment could not show relevant significant responses according to the resarch order, placed by Nammo Raufoss AS. As a concequence of this new variables had to be used in follow up runs.
According to high research costs the industrial experiment has not replications, only at the follow up run there was a triplcation of the center point ID. By a check with the biological results it could be shown how much information is lost by not do replication of each ID.
The biological experiment could be followed up based on significances, at least for one factor as requiret.
The industrial experiment has shown that the problem with derivations in roundness of the projectils was caused by a lathing process before hardening.
Key words: Fractional experimental design, analysis of contrasts, significance, design of follw-up runs, response surfaces, quality assurance of data- documentation, IS-ISO 9001 (EN 29 001) norm