Design & Factor Selection
Design Types & Categories
Dr Jiju Antony, Senior Teaching Fellow
Quality & Reliability International Manufacturing Centre, University of Warwick
Coventry, CV4 7AL, UK
An industrial experiment is generally performed to increase the knowledge of a particular process. One can improve the process performance only after an in depth understanding of the process. For example, in a certain injection molding process, the main concern for engineers was the parts shrinkage problem. An experiment will assist engineers under these circumstances to determine which of the factors (injection speed, injection pressure, mould temperature, type of resin, etc.) affect shrinkage the most. Design of Experiments (DOE) is a powerful technique to study the effect of these factors on shrinkage and also assists to determine the best settings of these factors to minimize shrinkage. The success of any industrial experiment depends on a number of key factors such as statistical skills, engineering skills, planning skills, communication skills, teamwork skills and so on. The DOE is a strategy of industrial experimentation so that reliable and valid conclusions can be drawn efficiently, effectively and economically. This paper presents some useful tips to industrial engineers for making their experiments in the work environment successful. It is important to bear in mind that these tips were developed strictly on the basis of experience in the topic and also by reviewing various successful industrial case studies in the area of study.
1. Good understanding of the problem
Research has shown that one of the key reasons for an industrial experiment to be unsuccessful is due to lack of understanding of the problem itself. The success of any industrially designed experiment will heavily rely on the nature of the problem at hand. The success of the experiment also requires team effort, which typically includes people from design, manufacturing, R & D, quality and most of all management commitment. In essence, obscure understanding of the problem often lead to lost time and money, as well as feelings of frustration for all involved.
2. Conduct a thorough and in-depth Brainstorming Session
The successful application of DOE in the modern industrial world requires a mixture of statistical, planning, engineering, communication and teamwork skills. Brainstorming must be treated as an integral part in the design of effective experiments. It is advised to consider the following key issues while conducting brainstorming session:
· Identification of the process variables, the number of levels of each process variable and other relevant information about the experiment
· Development of team spirit and positive attitude in order to assure greater participation of the team members.
· How well does the experiment simulate user environments?
· Who will do what and how?
· How quickly does the experimenter need to provide the results to the management?
3. Select the appropriate response or quality characteristic
A response in the context of industrial experiment is the performance characteristic of a product which is most critical to customers and often reflects the product quality (Antony,1997). It is important to choose and measure an appropriate response for the experiment. The following tips may be useful to engineers in selecting the quality characteristics for industrial experiments.
· Use responses that can be measured accurately.
· Use responses which are directly related to the energy transfer associated with the fundamental mechanism of the product or the process.
· Use responses which are complete, i.e., they should cover the input-output relationship for the product or the process.
It is not good practice to select attribute characteristics (i.e., good/bad, pass/fail, defective/non-defective) over variable measurements. One of the limitations with the attribute characteristic is its poor additivity property. It means that many main effects will be confounded with two-factor interactions or two-factor interactions will be confounded with other two-factor interactions. Moreover, attribute characteristics require a large number of samples and therefore experiments involving such characteristics are costly and time consuming.
4. Choose a suitable design for the experiment
The choice of design has an impact on the success of an industrial experiment as it depends on a various number of factors which include the nature of the problem at hand, the number of factors to be studied, resources available for the experiment, time needed to complete the experiment and the resolution of the design. The choice of an experimental design will be dependent upon the following factors:
· Number of factors and interactions (if any) to be studied
· Complexity of using each design
· Statistical validity and effectiveness of each design
· Ease of understanding and implementation
· Nature of the problem
· Cost and time constraints
5. Perform a screening experiment
A screening experiment is useful to reduce the number of process variables to a manageable number and thereby reduce the number of experimental runs and costs associated with the entire experimentation process. For example, one may be able to study seven factors using just eight experimental trials. It is advisable not to invest more than 25% of the experimental budget in the first phase of any experimentation such as screening. Having identified the key factors, the interactions among them can be studied using full or fractional factorial experiments (Box et al., 1978).
6. Randomize the experimental run, if possible
For industrial experiments, randomization is a process of performing experimental trials in a random order in which they are logically listed. It is generally recommended because an experimenter cannot always be certain that all important process variables affecting a response has been included and considered in the experiment. The purpose of randomization is to safeguard the experiment from the influence of lurking variables or noise, such as change of relative humidity, change of ambient temperature and so on. These changes, which often are time-related, can significantly influence the response. It is essential to quantify the effect of the overall background noise and then to reduce it to its acceptable limits prior to carrying out the actual experimentation (Verseput,1998).
7. Replicate each experimental trial condition (if possible)
It is important to note the difference between replication and repetition in the context of DOE. Replication is a process of running the experimental trials in a random manner. In contrast, repetition is a process of running the experimental trials under the same set up of machine parameters. In other words, the variation due to machine set up cannot be captured using repetition. Replication requires resetting of each trial condition and therefore the cost of the experiment and also the time taken to complete the experiment may be increased to some extent. Schmidt and Launsby provides a useful table for the number of samples (or number of replicates) required for an experiment with the aim of identifying a significant process variable or factor effect (Schmidt and Launsby,1992).
8. Use Blocking Strategy to increase the efficiency of experimentation
Blocking can be used to minimize experimental results being influenced by variations from shift-to-shift, day-to-day or machine-to-machine. The blocks can be batches of different shifts, different machines, raw materials and so on. Multi-variate charts (advocated by Shainin) could be a useful tool for identifying those variables which causes unwanted sources of variability (Bhote,1988). More information on the blocking strategy can be obtained from Box, et al. (Box, et al.,1978).
9. Understand the process using a sequence of smaller experiments
It is good practice to perform a sequence of smaller experiments at the beginning stage to understand the process behavior rather than trying to learn everything about a process from one large experiment (Antony,1996). If some initial assumptions go wrong (for example, the choice of response of interest) at any stage of the experiment, a significant waste and loss of management support may result. It is advocated to build up from two-level full or fractional factorial experiments in identifying the key variables early in the experimentation process. This should be followed by few more experiments to study the interactions among the key variables and also the presence of non-linearity of effects of the variables (if any) on the response.
10. Perform Confirmatory trials/experiments
It is necessary to perform a confirmatory experiment/trial to verify the results from the statistical analysis. Some of the possible causes for not achieving the objective of the experiment are:
· wrong choice of design for the experiment
· inappropriate choice of response for the experiment
· failure to identify the key process variables which affect the response
· inadequate measurement system for making measurements
· lack of statistical skills, and so on.
Product quality and process effectiveness can be accomplished in the modern industrial world by the use of carefully planned and designed industrial experiments. Both European and Western manufacturers have reported a number of successful industrial experiments. However research has shown that not many engineers are aware of industrial experiments for tackling manufacturing process and product quality control problems such as reducing scrap rate, quality costs, process variability, product development time and improving process yield, reliability and customer satisfaction. The author thinks it is quite important to classify quality and engineering problems based on their potential to benefit from the use of the industrial experiments. This is an area with a lot of potential for further research. This paper provides some practical tips to engineers for making industrial experiments successful in their own organizations.
Antony, J. (1996), "Likes and Dislikes of Taguchi Methods", Journal of Productivity, Vol. 37, No.3, pp. 477-481.
Antony, J. (1997), " Experiments in Quality", Journal of Manufacturing Engineer, IEE, Vol. 76, No.6, pp. 272-275.
Bhote, K.R. (1988), " DOE - The High Road to Quality", Management Review, pp. 27-33.
Box, G., Hunter, W. and Hunter, J.S. (1978), " Statistics for Experimenters", John Wiley and Sons, NY.
Schmidt, S.R. and Launsby, R.G. (1992), " Understanding Industrial Designed Experiments", Air Academy Press, Colorado Springs, Colorado.
Verseput, R. (1998), " DOE Requires Careful Planning", R & D Magazine, pp. 71-72.
Learn more about the DOE tools for designed experiments in Six Sigma Demystified (2011, McGraw-Hill) by Paul Keller, in his online Intro. to DOE short course (only $99) or online Advanced Topics in DOE short course (only $139), or his online Black Belt certification training course ($875).