Q 363. What is meant by "noise factors" in context of experimentation? How can one overcome the effects of these noise factors?

 

 

Note for website visitors - Two questions are asked every week on this platform. One on Tuesday and the other on Friday.

• All questions so far can be seen here - https://www.benchmarksixsigma.com/forum/lean-six-sigma-business-excellence-questions/
• Please visit the forum home page at https://www.benchmarksixsigma.com/forum/ to respond to the latest question open till the next Tuesday/ Friday evening 5 PM as per Indian Standard Time. Questions launched on Tuesdays are open till Friday and questions launched on Friday are open till Tuesday. 
• When you respond to this question, your answer will not be visible till it is reviewed. Only non-plagiarised (plagiarism below 5-10%) responses will be approved. If you have doubts about plagiarism please check your answer before submitting with a plagiarism checker tool like https://smallseotools.com/plagiarism-checker/ before submitting. 
• The best answer is always shown at the top among responses and the author finds honorable mention in our Business Excellence dictionary at https://www.benchmarksixsigma.com/forum/business-excellence-dictionary-glossary/ along with the related term

1. Uncontrollable factors that induce variation under normal operating conditions are referred to as "Noise Factors". These factors, such as multiple machines, multiple shifts, raw materials, humidity, etc., can be built into the experiment so that their variation doesn't get lumped into the unexplained, or experiment error. A key strength of Designed Experiments is the ability to determine factors and settings that minimize the effects of the uncontrollable factors.
2. Effects of Noise factors can be overcome by using Few techniques like  Blocking and/or Randomizations in the experiments.

Noise factors are concepts in Taguchi designs, which are inferred as  uncontrollable factors in normal operating environments which cause variations. However, during experiments the noise factors are controlled or modulated to emulate different conditions. Few examples of noise factors are - Environmental factors (external), Product deterioration (due to environmental exposure).

Noise factors are simulated in a controlled environment (for experiments) to test the product. There are also practices to compound different noise factors to produce extreme values and test the product robustness. 

 

The noise factor is an input process that cannot be controlled and this regularly causes variation in the output.
In Six Sigma Noise is commonly referred to as:
White noise
Random variation
Common cause variation
Noncontrollable variable

e.g. When riding a bike engine temperature depends on the combustion of fuel and the atmospheric temperature outside, here atmospheric temperature is a noise that cannot be controlled.

 

To control the noise first we have to assess it using methods such as ANOVA, DOE, or regression analysis.
Some parameters which are easily controllable in our system and which can interact with noise effects, then in those case we can use those control factors to minimize the noise.

 

There are 2 ways to reduce the variation or noise in our process:
1. Using non-linear effects - Response surface DOE can be used to learn curvatures and quadratic effect, and while experimenting we have to look for the sweet spot which has shallow slope.
2. Using Interaction effects - It's a interaction of noise factor with controllable factor. We have to look for    slopes and difference between the controllable factors, the lesser slope for control variables on noise indicates that it's handling noise better compared to other control variables.

Edited May 9, 20215 yr by Pankaj Goswami
Just added on how we can control the noise.

In Design of Experiments, we vary multiple Inputs (X's) at a time, in order to find the best combination of inputs to deliver the desired output (Y's).

 

During the same, we may come across some parameters which are uncontrollable. These are called Noise factors.

 

E.g. for an experiment on making the the right Dish, we can vary and control the quantities of constituents, temperature of cooking, time of cooking etc. But we cannot control factors like the mood of the Chef, the skills by which he mixes the dish etc. Mostly these human factors become noise and can affect the DOE results.

 

Noise factors are of 2 types : Known and Unknown

 

To address the same we can use 2 methods :

 

a) Known Noise Factors : Use Blocking (A block is a group of homogenous experimental observations).

 

b) Unknown Noise Factors : Use Randomization (make the Run order separate from Standard Order)

The noise factors are the design or process parameters those are difficult or expensive to control and affect the output of variable of key interest negatively. On the other hand control factors are controllable factors and can be controlled by the person doing the process or experiment. 

For example : 

i) Say a farmer wants to grow the wheat, then the seed quality, time of sowing the seeds etc. are control factors and can be controlled by the farmer; however on the other hand outside temprature, humidity etc. are the noise factors which are difficult or expensive to control.

ii) On the production floor, say itensity of the light impacts the productivity of the worker; then light intensity can be treated as the noise factor.

How to overcome the effects of noise factors ?

Effects of the noise factors can be reduced / eliminated by Blocking. We can divide the total population into homogenous groups called blocks. The logic behind making the blocks is the variation due to noise factors is less between the blocks and effect of the treatment is more clearly evident while we do the blocking. E.g. Say we want to calculate the productivity of the team and we think that shift timing is one of the noise factor that imapcts the team's productivity; so then we can calculate the team's productivity shift wise and address the problem effectively if it lies in a particular shift only.

Compounding noise factors is also a strategy in which you group the noise factor levels into different combinations that you anticipate will result into the extreme response values. Because estimating the effects of individual noise factors is not the primary goal, compounding is a useful method to reduce the amount of testing. For example, if you have three noise factors, each factor with the two levels, you will have eight different combinations of settings to test. Instead, you may group noise factors into two overall settings – one setting in which the noise factors levels increase the response value and the other one in which the noise factors levels decrease the response value.

Noise factor:

                   Noise factors are the functional characteristics that can be changed naturally, and it can be varies uncontrollably in the process. It is needed much effect to control while doing manufacturing process, but these noise factors can be controlled by the effective designs of experiments and these noise factors can be a measurable parameters/element.

 

Schematic Diagram for a Process is,

Firstly, the source of the noise factor can be identified by the detailed study of the system and it can be eliminated or minimized.

These noise factors affect the system performance adversely and the control variable have the capability to minimize the noise factors in the system. To defeat the effects of these noise factors, need to be analyzed and optimize the control factor which can reduce the noise effects.

During the experiment blocking and randomizations are used to control the noise factors and also the noise vs control factor interaction effects can be used to reduce the noise effects.

Noise factors in the context of manufacturing or experimentation are process or design parameters that are expensive, difficult or sometimes impossible to control. As opposed to Control Factors that are essentially process or design parameters that you can control. For example, while developing a printer the type and grade of paper the end user uses would be a noise factor. Or for a baking process the ambient temperature and humidity would be a noise factor.

Common types of noise factors are: environmental factors such as temperature and humidity, customer usage, part-to-part variations and Product deterioration.

 

A common way to overcome the effects of the noise factors is to force variability in noise factors during experimentation and then from the results of the experiment, identify optimal control factor settings that make the process or product resistant, or robust to variation from the noise factors.

 

In the above example of a printer development, during experimentation, the manufacturer could test several paper types to determine control factors that reduce the effect of paper type on printer performance.

 

In context of Experimentation, any "nuisance" or unwanted and uncontrollable factor which causes variability under normal operating conditions can be termed as a noise factor. These Noise factors may affect the dependent variable and hence adds undesired variation to the process / experiment and/or product. Such noise factors are typically of no interest to the researcher or process designer.

Most of the times Noise factors are difficult or expensive to control during manufacturing and hence, is best to be controlled and substantiated during the research/ experiment stage itself.

There are 4 types of imminent Noise factors-

1)    Physical Noise is any external stimuli in the environment that causes variation to the desired condition in an experiment or process. Common examples are someone talking incessantly in the background during an experiment where silence is required. This causes disturbance and is not required- similarly other examples can be unwanted music being played, dual communication etc.

2)    Physiological Noise is any unwanted stimuli caused due to physiological function like hunger, fatigue, headache etc. basically any physiological effect that affect the way you think and feel normally.

3)    Psychological Noise are mental interference like wandering thoughts, preconceived ideas

4)    Semantic Noise is any disturbance in transmission of any message due to misinterpretation that may be caused due to ambiguity in words, symbols, acronyms etc.

 

During an Experiment, the noise factor can be derived by taking the SNR (Signal to Noise Ratio) at the input and dividing it by the SNR at the output. The SNR at the output is always lower and hence it is obvious that the Noise factor is always greater than 1.  It is always better to have the signal level higher than the Noise level to have best output.

 

The above mentioned uncontrollable factors can be controlled during the experiment using steps like Blocking & Randomizing.

While Blocking is a process used to remove the effects of a few of many important or effective nuisance variables, Randomising is a process used to reduce the ill-effects of the remaining nuisance variables. It is obvious that blocking has higher significance than randomizing in reducing the variation caused due to impactful variables of interest.

 

Noise Factors in DOE are an uncontrollable factors which induces variation when operating in normal  condition. 
Various machineries, raw materials, temperature, humidity,  multiple working shifts, are some of the examples.
Taguchi’s methodology is an effective method to overcome the effects of noise factors.

All have given unique perspectives to their answers, however the two answers that stand out from the rest are that of Rahul Garg and Dipankar Acharya. 

Rajesh's answer is a must read to understand the different types of noises.