Difficult times like this require creative solutions to get the maximum performance out of production facilities. This requires optimum use of capabilities of the available resources material, equipment and people.
To improve the production results another resource may be of great help: historical production data. With the historical data and statistical modeling packages the production results can be dramatically improved.
In the past the statistical packages were very complex and expensive, but current developments in this area are showing simple configurable packages for reasonable prices. Process engineers with a little help from a specialist are able to built and maintain there own statistical models and can realize the possible improvements for their process within a number of weeks.
In many cases a process can be scientifically described by a number of formulas. However, in other cases, the exact working of the process cannot be explained in detail, since there is not enough knowledge about the process, the ingredients differ from batch to batch or the circumstances cannot exactly be controlled. In those cases a process model can be generated by statistical packages based on historical production data. The models can be used for troubleshooting and for on-line monitoring of the current batch.
The statistical model can be connected to a production control system. This allows for real time multivariable comparison of the actual batch with the Golden batch.
The multivariable trend indicates that the current batch deviates from the golden batch at the beginning of the process. At the same time the statistical package indicates the most likely reason for this deviate. In the diagram below the package indicates that the pH is causing the change with a confidence level of xx%
In many processes a business case can be set up to demonstrate the profitability of using a statistical package to optimize the process. In most cases required investments are less than € 100,000 and have a payback time of 3 months or less.