My Dear Sreeja,

Your question is difficult to answer. How ever i will try to give some opinions to it.

At first one has to define the errors of human concerning certain activity. This can be accomplished by determining the right expected performance and then measuring the real performance and the deviation is the error. For example shooting a target by a pullet using a gun. You see that the probability to hit the center of the target is less than one because random human errors or random environmental conditions such as the speed of wind and its direction.

In fact the deviations from an expected value or behavior is because of the sensitivity of this value to statistical factors called tolerances. In fabricating any product one finds always that this product deviates from its standard values and this deviation is normally distributed around certain most probable value or the average value. When the human intervenes and share in the production process normally one find that the deviations increases because the human adds also random factors to the production because of his tolerant behavior.

Therefore the production processes are fully automated to exclude the tolerant behavior of the humans.

Best wishes

Thank you very much for your insightful answer, Dr. Zekry.

I think these references can be useful for you:

https://inis.iaea.org/collection/NCLCollectionStore/_Public/38/115/38115083.pdf

http://assets.newport.com/webDocuments-EN/images/Solar_Industry-Solar_Cell_Testing.PDF

I too find this to be a difficult question. The errors usually handled by descriptive statistics such as sample standard deviation are, by definition, indeterminant- meaning they cannot be assigned to specific errors, including "human errors." Determinant errors would affect the mean, but should not impact the standard deviation and may reflect procedural problems. However, comparison of the standard deviation (or confidence intervals or tolerances) achieved by an individual analyst or producer, to that achieved by typical experienced analysts certainly indicates poor execution, e.g. misidentification of the true endpoint of a titration when using an indicator. As stated above, this has been one of the main drivers of automation. In the case of human operators, any standard operating procedure must be very specific while being easily understood.

Thank you very much for your insightful answer, Dr. Coolbaugh and thank you for the documents shared, Dr. Asgari. I'm having having trouble quantifying the error while fabricating solar cell configurations that do not have standardized data - i.e. they are made of new materials. In this case how do you think one should go about it?

The document Dr. Asgari has shared addresses the error in measurement however does not really address the error in fabrication. I understand that it is difficult to quantify but was wondering if anyone has done that in a systematic way.

Dear Sreeja,

I think your question is missleading and do not directly describe your problem.You must be specific and go to the real problem instead of speaking generally on human errors which may be countless and diverse.

You have a structure which can operate as a solar cell.

-The first step is that you have to be knowledgeable enough to qualitatively analyse the new structure and predict its expected performance.

- Then you can use the CAD tools to quantitatively analyse the structure and get out its optimum structure.

- After getting the best theoretical structure you go then to the experimental work

- You have to plan your experimental work and define the fabrication processes elaborately

- You have to avoid any effects of the environmental conditions on your experiment for example the contamination of materials from the environment.

- After every fabrication steps you conduct the required tests to be sure of the success of this step.

- Fabricating the device as a whole and waiting to test it at the end of the fabrication can cost you more time to trouble shoot the malfunction of the structure.

- Of course you have to study the previous similar or related works and consider its lessons to save the number of trials to reach your target.

I think the main error of the human is to not to make preliminary study and plans for his work

Best wishes

Thank you very much. That's very helpful.

I can't resist trying to add to this discussion even though I should be getting ready for class. I will try to add more later, but I think you may want to think about applying Design of Experiments methods to your process. No matter what, you need to carefully consider which performance characteristics (responses) you want to optimize and exactly which parameters in your process you can precisely control (factors.) If you have factors you cannot control, then I fear you will never achieve any meaningful success in your endeavor.

Even though factors have been controlled at every stage as much as humanly possible, the efficiency values achieved for a given set of samples of identical configuration are never exactly repeatable and always fall within a range (e.g. 1% +/- 0.06%). By how much is this deviation considered acceptable?

I think it would be a bit more useful to actually see the descriptive statistics here- number of runs, the standard deviation and perhaps the 95% or 99% confidence intervals. Simply stating the range of values is doesn't really give us much idea of how the process is working. There are some reasonably good statitical packages availabe, but if you have access to Excel you can easily perform basic statistics.