Optimization of experimental Procedures using Design of Experiment / pdf 

 

 

Experiment precision and repeatability was improved using DOE and GR&R. 

 

The Case and Objective

The measurement of an organic content determination experiment was giving inconsistent results both within one site and among different laboratories.

The objective of the project was to determine the bias, delta between the measurement and real value, and to standardize the measurement protocol in all sites of the organization in accordance with CE. Basically, it has two steps that at first the water is evaporated and then the sample calcinated to burn organics. Critics about LOI experiment can be marshaled in clauses.

• Low-cost experiment

• 3-hour procedure

• High variation (±10%)

• Health and safety care is essential, - not to inhale the organics calcinated 

 

 

 

The Work

Heretofore, the experiment bias was unknown. It is prepared a bias experiment set with the help of design of experiment tool and the reproducibility and repeatability of experiment examined using gauge R&R study.

The logic behind bias study was to prepare samples having pre-determined value of organic content. The brainstorming sessions made to find out a manner to prepare reference samples. The corresponding amount of binder is sprayed to the samples to achieve a predetermined known value.

Eventually, we know exactly actual value of each sample that is the reference for measured value to compare. The best experimental conditions were determined with 3 factors and 2 levels. The very basic idea behind this set of experiments is that to compare the known values with the measured ones. Delta bias study shows the consistency and magnitude of delta between the reference value and the measured one.

 

In bias study output the best fit to the reference is to be parallel to horizontal line and close to zero. As the current case might be seen in the graph above the bias is not parallel and not close to zero, which interpreted as very much delta and inconsistent delta between measured value and the real reference value, variation decreased from 10% to 1%.

 

In another setting of DOE, it has been proved that bias is almost zero and the consistent, as in the following graph. 

 

Furthermore, the GRR study made to determine the repeatability of the experiment by same operator for many times and reproducibility by different operators. Non-of the points in R-chart (range chart) are outside the control limits, indicating the stability of the process.

 

Most of the points in X-bar chart, (Means Chart) almost all points are outside of the control limits, indicating that variation is mainly due to differences between parts. This means that part variation exceeds measurement system variation. Repeatability of the experiment decreased to 37% at 2% process tolerance.

 

 

The Result

The experiment bias was determined and minimized and reached to consistent measurement. Moreover, the reproducibility and repeatability of the experiment is determined and improved thanks to new settings.

This new method was totally harmonized with CE, ASTM and AENOR standards. The harmonization is announced to all sites of the organization to measure in a standard way with a dedicated SOP. The standard operating procedure and systematic work created thanks to this strong baseline project substituted the previous doubtful case.