|
|
DOE - Article
|
Case Study
|
Guidesheet titled:
Design of Experiments; Improving Your Product and Processes
(Level: Basic)
|
Company background:
A locally-owned electronic manufacturing company with plants in Pulau Pinang.
|
Products description:
Telecommunication products such as surface mount technology (SMT) circuit boards, printed circuit boards, speakers, microphones, modems, etc.
|
Main markets:
Europe and United States of America
|
During our three-month stay there, several DOE techniques were identified as being relevant to the company, including factorial designs and fractional factorial designs. When there are many factors or variables to study, fractional factorial designs are more appropriate than factorial designs. With the help of the company's senior QA manager, we identified the SMT process as one of the possible areas for the application of these two techniques. Basically, this process is made up of the following four sub-areas; screen printing, glue dispensing, chip mounting, and reflow curing. We decided to focus first on screen printing since this is first and most critical of the four sub-areas.
A quality problem that can occur at screen printing is excessive variability in the height of the solder paste deposits. Not only can this occur from site to site on each solder pad, but also from solder pad to solder pad on each circuit board, and from circuit board to circuit board in each lot. Besides wanting to minimize this variability, the company would also like the solder height to be at the nominal specification limit. A team was formed, comprising of SMT production engineers, QA engineers, QA section manager and us. The objective of the team was to go over the actual mechanics of applying the DOE technique of factorial designs and fractional factorial designs to screen printing. The first step was to use brainstorming technique in order to identify the variables that could effect the solder height such as squeegee speed, squeegee pressure, stencil thickness, etc.
The question that were of interest to the team members were then formulated as follows: which of the identified variables were the ones that actually affected the height, what should their setting be, and what actions(s) should be taken based on the answers to the first two questions. Among the things that were then discussed were the range of acceptable values for each of the identified variables, the number of runs to perform, the details of each of the runs, the number of circuit boards to use for each of the runs, the location of the solder pads on the circuit boards and the location of the sites on the solder pads for taking the solder height measurements. The repeatibility of the solder height measuring instrument was also investigated.
Training session on how to plan experiments using the above designs, and how to analyze the results of the experiments using both graphical and statistical methods were then held for the company's engineers and managers. The training session also included the use of Taguchi Methods; which is a modified form of fractional factorial designs.
After the training session, the team use the DOE technique known as a full 2 4 factorial design on the screen printing process. Four variables identified as A, B, C, and D in this guidesheet (the actual names of the variables are not revealed here for confidentiality reasons) were investigated using 16 runs. The variable were varied over two different levels; a low level and high level. All other variables were held fixed at a particular level. For each of the runs, the height of the solder paste deposits was measured from three circuit boards, and at 10 randomly selected sites on each board.
The measurements from the 16 runs were analyzed using graphical and formal statistical methods. The variables that were found to have a great influence on the mean of the solder heights were variables A and C, and the interaction between variables A and C. Variability in the solder height was greatly influenced by variable D. In order to obtain the required solder height value (i.e. solder height near the nominal specification limit) with minimal variability, the following settings were recommended; high level of variable A, high level of variable B, low level of variable C, and low level of variable D.
Our next set of guidesheets titled ' Understanding Factorial Designs' focuses specially on the planning and analysis aspects of factorial designs, and touches on fractional factorial designs.
|
|
|
|
