![[Zimaths article logo]](artic.gif)
Click here for Part I of this article
Right! Our stocks are sufficient and have been assessed for quality. A large order has come in and we are ready to start production. How can Statistics help us to produce a tasty confection? Keep reading.
What happens on the factory floor? Simply put:
Then, finally, the bar is ready for cutting, packaging and marketing.
In order for ingredients to be accurately measured, temperatures to be set correctly, pressure to be maintained,...., our machines must be in fine working order. Using calibration techniques this is straightforward and reliable. Calibration can be likened to the reverse of regression i.e. the fitting of a model to a given set of parameters. Tolerance levels can also be investigated in this manner.
At each stage on the production line one must test for quality. Control charts are expressly designed for this purpose as they keep track of the process. Any deviation away from the norm is immediately detectable and can be acted upon with minimum time loss. So for example we can maintain a control chart for temperature, which should not vary by more than 1 degree to avoid spoiling. Any trend, an upwards or downwards shift, will be noticed and remedies taken to correct it, before a critical stage, resulting in loss of the product, is reached.
Process capability is the study whereby we ensure we are working at maximum productivity whilst remaining within the laid down rules specifying the end quality of a product. For example, a bar of chocolate has to have a certain minimum weight; it is not economically sound to produce bars which are too heavy so as to ensure they do not fall below those limits. The probability distribution of the weights will give us confidence levels within which we can operate with pre-set levels of probability of success.
How reliable are all these measurements we take? Do they really represent what is happening in the vats of cooking chocolate? Can they predict our outcomes?
In order to monitor efficiently one should select items for inspection as they pass through crucial stages of the manufacturing process. Random selection is essential in order to gain a true picture of the intricacies of the many processes combining to produce a single item. Thus for example we would sample the item halfway through the process and rush it to the laboratory for testing. The results will determine whether it is ready to be moved through to the next stage of production.
Sometimes one wishes to trace the progress of a particular raw material through the production stage. An example of the usage of this would be if we wished to use a new and cheaper brand of milk powder (a necessary ingredient of chocolate) but are unsure as to how it will stand up to the manufacturing processes. By tracing the progress of the new powder during the manufacturing we can monitor its suitability.
We're at the end of the line! Samples have been thoroughly tested and the laboratory tells us we can move forward and proceed to packaging. Find this spot in the next issue for the next exciting instalment!!