Batch control makes chemical reactions easier to manage

Two Dutch researchers have developed a method for managing so-called batch productions. During a batch production, substances react in a reactor vessel according to a certain recipe to produce an end product. After the reaction the reactor is emptied and a new reaction with the same recipe is started.

Chemist Eric van Sprang and chemical engineer Henk-Jan Ramaker have developed a control method that also takes the relationship between various process parameters into account. The current methods of process control monitor all of the parameters during the reaction, such as pressure and temperature, separately. As a result of this the control process costs a lot of time and not all of the process disruptions are clearly visible.

Batch control is important for safety, the environment and product quality. The quality of the product made in a batch process depends on the various parameters involved in the chemical reaction. However, these parameters are never the same for all batches.

The researchers made a model to predict how large the variations can be without endangering the quality of the product. They first of all collected the process parameters from more than thirty batches and then described the process variation with the help of a model. Finally, they used this model to make two control cards that an operator can use to control the process. If the process parameters of the reaction remain within the control limits, the process is proceeding as intended. If that is not the case, there is a process disruption.

Traditionally factories control a chemical process by monitoring several parameters such as pressure and temperature. These are measured by sensors in the reactor. The outcome of each separate measurement is noted on a so-called univariate control chart. Therefore several charts are needed to monitor several variables and this means that the relationships between different parameters are ignored. For example, if the pressure in the reactor vessel increases, the temperature often rises as well. However, if this relationship no longer holds due to a process disruption, univariate control charts might not detect this.

However, the new method from Sprang and Ramaker takes the relationships between process parameters into account. With this method the process operator in the control room only needs to monitor two control cards. The early and reliable detection of process disruptions leads to indirect cost savings on the process. Therefore it is essential to choose as good a model as possible in combination with the correct statistics.

The research was funded by the Netherlands Organisation for Scientific Research.