The Test Plan #1
Contents
2. The Test Plan #1¶
When planning an experiment we need to keep in mind the main question it was deigned to answer. In many cases, the answer we seek will depend on more than one variable, and not all of them will be controllable.
Before starting to design our experimental system, we must gain a good understanding on the process we aim to measure, the theory behind it and the variables that may influence it. We will cover some of those more thoroughly in section G.
2.1. Think, plan, think again and experiment¶
In general, when constructing our test plan, we will need to consider several principal issues:
The question - experiments are aimed at answering a question. properly defining our goal is the first step to fulfill before carrying out any further work. The question may be phrased as an hypothesis “If A than B”; as a query “How A will affect B? what is the value of X?” or as an optimization problem “what is the best set of \(Y_i\) to improve \(Z\)? “. The question we phrase will motivate the entire experimental approach. It will be key in answering further questions down the road such as “to what extent do we need to control the environmental inputs?” ; “What is the required accuracy” ; “What levels of uncertainties are we willing to tolerate?”.
Measurement system design - Based on the question we phrased we need to ensure that the measurement technique will yield the data we are looking for under the required tolerances. Skipping the first step in the design stage, we may come to realize that we have ignored crucial aspects of the problem such as humidity, the order at which the experiments were held, the operator proficiency and many more.
Data analysis - what tools will we use for analyzing the results? what will we accept as a successful experiment? what kind of comparisons can we draw to validate our experimental approach and data? While this last step may seem as something that can be postponed to after we obtain the experimental data, it will often make us rethink our entire experimental approach and will send us back to step 2 or even step 1. As such it is highly recommended not to skip it.
2.2. Basic concepts for designing experiments¶
Process - the transformation of inputs (controllable or not) to outputs. The inputs or factors may be quantitative (e.g. velocity; test temperature; current level etc.) or qualitative (material supplier; operator etc.). for each factor influencing our process we will need to decide on the number and range of levels we will measure and/or control. When possible, we will ask ourselves if we should group variables together ad parameters or rather study their individual effects.
Interaction - two factors are said to interact if the effect of one factor on the other is dependent on the level of the other factor.
Replication - In the majority of experimental systems, some variation between repeated experiments will always occur. Replication refers to the repetition of the experiments under more than one condition. The importance of repetition lies in its ability to help us estimate and reduce experimental errors as well as refining the observations regarding interactions effects.
Replication should no be confused with Repetition we will repeat the experiment under a given set of conditions as dictated by our experimental plan to reduce the experimental measurement error. Replication will be used to estimate the effects of external or other unaccounted for factors.
Randomization - Factors such as environmental conditions, the amount of time a machine was used continuously, mechanical wear, electrical noise etc. are inevitable. Randomization of the experimental plan will aid in reducing the bias caused by those factors, by giving all levels of a factor an equal chance of being affected by them.
Note
This is just a partial list of basic concepts. We will return to the Design Of Experiments (DOE) later in the course.