Nutek, Inc. Quality Engineering Seminar, Software, and Consulting ( Since 1987) You are at Seminar site |
HOW TO GUIDE for |
Planning Tasks
Nutek, Inc. All Rights Reserved. |
||
About this HOW TO Guide |
||
Proper planning is key to success of your experiments. You must plan the experiment as best as you can with as many people in your team as practicable and follow the structured path of discussions prescribed in the steps described below. This HOW TO steps guides you through the sequence of discussions and help you collect information vital for the success of your experiment. In sections below, wherever applicable, the tasks you are to do in the step are being described as bulleted items. If you are familiar with the activities, this items list will serve as a quick reminder. As you follow the list, collect your thoughts and/or develop group consensus, be sure to manually note what you decide (Title, QC, Factors & levels, etc) in a separate document. You will need this information when you move on to the experiment DESIGN tasks using Qualitek-4 software after this planning process. |
1 |
Forming Experiment Planning Team for Brainstorming |
|
Your project team (3 - 12 people) should consist of project stakeholders and people with first hand knowledge about the project. You should attempt to form a team and proceed with experiment planning following the guidelines provided below. Carrying out an experiment is same as completing a project. Like a project, your experimental effort will have a team leader and project team. For coordinating discussions and information gathering, your experimental project will also benefit from an independent facilitator. Of course, for small projects, you or project responsible individual can play all three roles (leader, team, and facilitator). Regardless of the project size and whether different individuals perform these roles or not, it would be helpful to carefully examine the tasks involved in these roles. Identify and fill these three positions based on the roles and tasks they need to accomplish:
Your experiment planning (or project) team should be consulted on selecting a day and location for the planning meeting. You should secure commitment from all team members for the entire day (8 AM – 5 PM). For efficient use of time, you may breakdown your project team in two different groups. Project team – Group 1: Team leader and few key people in the project with responsibilities to make decision about the project goals, objectives and implementation. It should include management personnel (stakeholders) with direct responsibilities about the product/process under study. Project team – Group 2: Team leader and members with direct knowledge about the project who will have clear understanding about how to set up the tests and limitations of the variables to be included in the study. On planning day, if you are the leader as well as the facilitator, you must be extra caring and conduct the meeting in most professional manner (fair and objective). Seek out information of the following listed items by involving everyone and by consensus when appropriate. |
2 |
Selecting Project Title |
|
Your project title is an important identifier of the activities you are undertaking. It needs to be something that relate to the product or process under study. It should include works that indicate the purpose of the study. Thus if your application involves studying a process that makes a plastic bracket to determine better process parameter, your title may be “Plastic Molding Process Optimization Study”. Since the purpose often is to solve problem or optimize designs, your title should include words like STUDY, INVESTIGATION, EXAMINATION, etc. To help you and your team think through the process under study, here are some questions for which you are seeking answers.
(Research, Design, Validation/Test, Manufacturing, Fulfillment/Deliver, Product Support, Warranty/Loss Prevention)
i. process optimization ii. product design optimization iii. problem investigation and solving iv. Validation test layout.)
i. What is it that is considered to be creating problem? ii. What will be absent when you solve this problem?
|
3 |
Describing Objectives of Experiment |
|
Your reason for the experimental study may be to solve a problem, optimize designs, layout validation tests, or increase response from advertisement, etc. No matter the real reason, you may view the activity as trying to solve a problem that is to fill-in a void or absence of something. In other words, when you are finished with the experiment, you will obtain something that you do not have now (problem). While stating the problem, indicate the benefit you will derive as a justification for doing the project. Here is how the project description can be composed for the PLASTIC MOLDING project indicated earlier. Example DESCRIPTION: “We have been experiencing high rejects and warranty from our plastic molding process. This study is undertaken to determine process parameters that will reduce our scrap rate. The improved process design is expected to keep our customers satisfied and affect our bottom line.” Here are a few questions that will help you collect your thoughts and determine suitable answers. - Your reasons for performing this project are? - What is it that you want to accomplish with this project? - List your objectives/goals for this project In case your study involves BAKING POUND CAKES, you may consider the objectives to be to: (a) Improve TASTE, (b) Increase MOISTNESS, (c) Prolong SHELF-LIFE. |
||
4 | ||
In your study, when you run a set of experiments laid out by DOE, you would want to see how each test sample performs in terms of satisfying the objectives. But, how would you know how well a particular objective is satisfied? Each objective can be evaluated by one or more characteristics called the EVALUATION CRITERION which measures how well an objective is satisfied. For example, an evaluation criterion like “a subjective scale in the range of 0 – 10” can be used to evaluate TASTE of cake. Where as, the criterion “weight in grams” could be used to evaluate MOISTNESS of cake. Consider the case of a study involving heat treatment of a steel part in which one of the objectives was to improve strength. The criterion of evaluation for the strength, in this case, was “tensile stress”. (Just identify the name. The units of measure and range for these criteria of evaluations will come later). Repeat this process for all objectives. Helpful task and question:
|
5 | ||
The criteria of evaluations you use to measure how well each objective is satisfied may be one that is a standard practice in the industry, or defined uniquely for your results. They may also be subjective or objective. If any standard unit of measure for a criterion is absent, you will need to establish the units of measure (lbs, inches, grams, numeric, etc) and its expected range (example: 0 – 8 for numeric, 200 – 300 psi for pressure, etc.). Consider each criterion and determine their units and range. Repeat this process for all criteria. Remember, all criteria of evaluations must be in quantitative terms. Helpful tasks and questions:
i. If Numeric = “What is the unit of measurement? [examples: temperature, %F or %C; concentration, %; quantity (any number)] 1. Establish range (low & high values) for each of these measures that you expect the experimental results to fall within. ii. For analysis of DOE results, you must have a numerical value for evaluations. For subjective evaluations, select a value scale that could be used for measurement?” Options given: 0-to-5, 0-to-10, 0-to-100, etc. Select higher ranges (0 – 20, 0 – 100, etc) only when you are able to discriminate between two successive numerical evaluations (say between 19 & 20)
|
6 | ||
Having identified all objectives and their respective criterion of evaluation, you are in a position to determine their individual quality characteristic (QC) and compile the information in the following table. Description of Criterion | Worst Value | Best Value | QC | Rel. Weight 1. Taste 0 10 B 2. Etc. After all criteria are listed and their QC identified, you will proceed to prioritize their importance in terms of relative weight. You may easily do so by allowing each member of your project team (also true when you are alone in the team) to distribute 100 pennies ($1) among all the criteria. In doing so, one must distribute all (100 pennies) to the criteria as per the importance in his/her own mind. After all team members are done assigning pennies to the criteria, add all pennies assigned to a criterion and divide it by the number of participants. The number you obtain, represents the group consensus on relative importance of the criterion. (Make sure the sum of percentage weights of all criteria always add up to 100) Example: Cake Baking Process Study Description of Criterion | Worst Value | Best Value | QC | Rel. Weight (%) 1. Taste 0 10 B 55 2. Moistness 25 gm 50gm N 30 3. Smoothness (#voids) 6 0 S 15 When you have multiple objectives, a common practice will be to analyze results under each criterion separately. Such separate analyses are then expected then to produce different factor influences and optimum conditions. It would be a matter of extreme coincidence if all optimum conditions turn out to be the same. As an improvement over this practice, you may wish to combine results under all evaluation criteria into a single index, and then analyze using this index. The method for combining such multiple objective results in to an overall evaluation criterion (OEC) is described the texts by R. Roy (http://nutek-us.com/wp-txt.html ). For quick refresher on the theory and background, you may visit: http://nutek-us.com/wp-oec.html . Should you want to see how Qualitek-4 handles such results of multiple criteria, select example experiment POUND.Q4W and use OEC Results from EDIT menu and review the method of combining different evaluations into OEC for each test sample.
|
8 | ||
|
|
The list of factors (A, B, C, etc.) in the Study List now needs to be carefully examined and their levels determined. The first issue in determining the level is to decide how many levels this factor should have. Generally, all factors should have two levels, but may have three or four levels. If a factor is a discrete/fixed factor (like tools, machine, shifts of day, male/female operator, etc.) it may have more than two levels. Also, if a factor is KNOWN to have nonlinear behavior, it may be necessary to study it in three or four levels. Otherwise, you should study all factors at two levels when possible. Approach you should follow is to consider each factor separately and determine:
Use these guidelines:
This way, complete determining the levels of all factors in the study list. Should you have a factor that needs to be at more than 2 levels (3 or 4), you will then need to drop factors to make room for this factor (level upgrade) or go for a larger array. For instance, if you happen to have one of the 7 factors to have 4 levels, you will need to drop 2 of the seven factors to make room for a 4-level factor. You will then modify the L-8 array to accommodate this 4-level factor and four remaining 2-level factors. However, before you can complete the design process, you will need to consider INTERACTION and NOISE factors that might be part of your study and may indeed reshape your experiment.
|
9 |
|
|
|
|
Now that you have selected the factors for the study and determined their levels, you need to consider possible interactions that you may want to include in the study. For interactions, consider only the interactions between two 2-level factors (like AxB, BxC, etc.). Understand that, if you have 7 2-level factors, there are possible 7 x (7-1) = 21 interactions. You are now faced with two questions: how many interactions to study, and which ones among all possible ones to study. Generally, you do not have any knowledge to answer these questions. But, if you happen to have the knowledge and/or conviction to decide upon some interactions to study, you will have to revise your experiment design. Suppose that you have two interactions that you must study. Since your limit on the size of the experiment is 7 columns (in an L-8), you could do so by discarding two factors to make room for he two interactions that have a common factor. If on the other hand, you want to study all 21 interactions and 7 factors, you will require an array that has 28 or more columns. To do so, you will need to increase the size of the experiment and go for an L-32 array. A general recommendation is that, you select the biggest array possible and accommodate all factors first. Then if you have spare columns, reserve them to study interactions. The last item to consider before finalizing experiment design is to consider the possibilities of formally incorporating the effects of noise factors in your experiments. The most desirable way to include uncontrollable/noise factors in your design is to go for an outer array design where an orthogonal array is used to formally combine the noise factors to create some critical conditions. To select robust design conditions, the tests under different recipes of the control factors are tested by exposing them to the influence of the noise condition so created. The noise factors, of course, are uncontrollable in real life, but are assumed to be controllable while conducting the tests under laboratory environment. If there are 3 noise factors in your experiment, you would use an L-4 array as an outer array. This will require that you run each trial condition (of the control factor) 4 times by exposing them to 4 separate noise conditions. Such formal treatment of the noise factors, require more samples and time in carrying out the experiments, but is likely to produce more useful information about the system under study. A general guideline to follow is to go for robust design approach using outer array, if not possible, carry out multiple sample tests in each trail conditions under random noise condition. |
10 |
|
|
|
|
Before you complete experiment planning discussions, you should know and be able to tell your team about the scopes of the experiments in terms of how many test samples would be involved and how the results should be collected. You could accomplish this by simply creating a table of data collection sheet that shows how many tests (trial conditions) would be done and what results would be collected under how many samples and criteria of evaluations. |
11 |
|
|
|
|
Before you adjourn your planning meeting, you need to discuss and share plans for acquiring test samples, test facilities, and data collection procedures with all in the team. If possible, you should also form consensus on the length of time and schedule of completing the study. Prepare a summary of the information gathered from the planning session. This could be a quick review with the group before you adjourn meeting with the team, or prepare it after the meeting and share it with the team members. You will need this summary page when you start using Qualitek-4 to design the experiment. Your planning summary should contain the following information. Project Title ________________________ Location _____________________ Participants: 1.__________________ 2.____________________________ 3.__________________ 4.____________________________ Criteria Description Worst Value Best Value QC Rel. Weighting 1. 2. 3. etc._____________________________________________________________ Your OEC Equation (if planned) OEC = ( ) x + ( ) x + ( ) x + ( ) x Example: FACTORS Level 1 Level 2 Level 3 Level 4 1. 2. 3. etc._________________________________________________________ List Interactions and Noise factors you wish to include in your study. Proposed Experiment Design: Indicate the inner and outer arrays used for the experiment design and how the control factors and noise factors will be assigned to the columns of the arrays. Based on the proposed design, indicate the test sample size requirements.
|