Plan and Run Experiments

Plan and Run Experiments

In the previous lesson we talked about Establishing Experimental Reproducibility.

Now, we will talk about how to plan your experiments.

Imagine you are planning a road trip. You are going to drive from Cleveland to Washington D.C.  You can get a set of directions from Google Maps, MapQuest or even a GPS system. Sometimes, the old fashioned map is better. You open an interstate road map map. Locate the starting point; say Cleveland, my hometown, and then the destination, Washington, D.C. You identify the shortest route, the fastest route and sometimes a few alternatives. Then, you select one that makes the most sense. Remember the road map analogy. It is a very useful analogy.

Planning Experiments, or more precisely and technically speaking, designing experiments, is similar. Architects use both of these words for designing a house..

      Planning refers to scheduling,

      Design refers to house layout, exterior look and so on.

Since we have already completed Step 2: Establishing Reproducibility, we already know the starting point. What performance we are getting today with the current experimental conditions?

We know the destination, the goal. See “Lesson 2. Define the Project” If you need a refresher.

However, we have a problem. We cannot use the road map. The road map does not exist.

So now we have a bigger challenge. We have to build the road map.

The road maps are drawn on a piece of paper. Typically in a horizontal direction, left to right is West to East; the vertical direction, from top to bottom, is North to South. The road map is two dimensional. We have a much bigger problem. Our problem is multidimensional. We want to study several variables (eight for the cheesecake problem – Lesson 2). We will also need a map for each performance characteristic we are interested in. We can lay all these property maps side by side and identify the right combination.

Actually it is much worse.

The cheesecake map is multidimensional. You see we humans can only see in two or three dimensions, but not in four or more dimensions!

Imagine we are blind! Where am I? How do I go to my destination? Which road should I take? Where is that road?

Now you know why it is so difficult to develop new products/processes.

Here is the good news. FReD, the bot, will make our job an order of magnitude easier.  

Just upload two files,

  1. Project Description
  2. Project Definition

and see how FreD, our bot,  helps us out. We will cover the "Why" in detail later. 

Figure below shows the default layout when you enter the DOE page. The default layout shows:

  • The Slider for Number of Control Variables is set to "3."
  • The Slider for Number of Expts, short for Experiments, is set to "8" 
  • The default list of experiments shows four "baseline" experiments at the "center" of the experimental region specified through Project Definition. 

We need to make two simple changes:

  1. Select the Number of Control Variables as "8.
  2. Select the number of Experiments as "16." We recommend a number multiple of 4.

You can "Download" a copy of the list of "Designed" Experiments as an Excel CSV file for ready use. 

You are ready to run these experiments in the Lab. I recommend:

  1. Start, with the four Baseline Experiments. (see Establish Reproducibility)
  2. Randomize the 16 experiments.
  3. Run them in the lab.

This concludes Step 3 of the DEPLOY Process, Plan and Run Experiments.

About Author

Mukul Mehta

Mukul Mehta has over 40 years of proven industrial experience in chemical , polymer, and plastics industry. Worked as a Sr. Manager, Statistics and Computer Aided Research for BF Goodrich Chemical, a Fortune 500 company, and then as a software entrepreneur, promoted "quantitative, predictive modeling in one minute or less as a mantra for R&D and New Product Development." Many multi-million dollar successes for dozens of Corporate R&D clients in chemical and pharma industry. Trained over 750 R&D chemists, engineers and managers to Speedup New Product Development through statistical design of experiments.

{"email":"Email address invalid","url":"Website address invalid","required":"Required field missing"}