Shortening preparation time and reducing downtime in the pasteurization/sterilization process by optimizing control loop design

The pasteurization and sterilization processes are among the most important unit operations in food processing. They consist of heating the food to high temperature for a predefined length of time. Those two processes aim to reduce the number of viable pathogens, so they are unlikely to cause disease. Compared to pasteurization, sterilization will bring the food to higher temperature which will kill all micro-organisms and allow the food to be kept for a longer time.

Both processes also consume a significant amount of energy in the form of electricity or fuel (steam production). They can have a big economic impact on the final product cost. In fact, the standard is very strict and production tightly controlled by individual country organizations (eg: USDA for milk in USA). Being over the temperature will alter the product quality, flavour and may accelerate clogging the pipes. Being under will require the product to be re-circulated and the whole production facility to be re-sterilized.

The pasteurization and sterilization processes are represented by the following drawing:loop pic1

The raw milk is pushed to the heat exchanger to raise the process temperature requested (eg 138°C for Ultra High Temperature sterilization treatment, 72~75°C for pasteurization) this temperature is then kept for the requested time during its circulation in the holding tube. The milk is then cooled in the cold exchanger. The result is pasteurized or UHT milk.

Usually, the heating temperature is controlled by a PID control loop that results in a setpoint sent to the control valve which manages the flow of a hot fluid. The heating fluid may be a liquid or steam. It depends on the pasteurizer design and the temperature required.

The cooling temperature is also controlled by a PID control loop by sending a setpoint to a control valve which manages the cooling fluid flow in the exchanger.

As previously explained, the biggest challenge is on the heat control loop. There are several disturbances in the process that may set the milk temperature out of range. These disturbances are:

  • Product viscosity variation. The process is starts with water and then switches to milk. The exchange coefficient of both is different and requires the heat controller to react quickly.
  • Product flow variation. Tank replacement, lack of milk.
  • Steam pressure variation. If steam is coming from a utility shared between several processes, the pressure may vary depending on the quantity of steam consumed by other processes at the same time.

I have been working recently with a dairy to help them optimize their sterilization process.

The trend below shows the sterilization start phase when the system starts to push the water in order to reach the configured temperature setpoint (here, 140°C). The milk will be pushed only when the water temperature stabilizes at the setpoint value.Temp1

The temperature is represented in red, the valve output in yellow and the steam pressure in purple. We can observe that the temperature increased to 140°C then it incurs an overshoot of 3°C and a small oscillation.

The same overshoot appears when the water is replaced by milk which is pushed at twice the flow speed.

On the trend we can see that without any particular disturbance we already have an overshoot of 3°C. If at the same time we had a disturbance, our results would have fallen outside the 5% percent limit (+-8°C) authorized by the process. It would result in the need to re-sterilize the entire production chain (which takes about 30 minutes). In addition, the time to reach 90% of the sterilization setpoint (140°C) took 13 minutes.

If we decide to slow down the controller to reduce oscillation, in this case the controller becomes too slow to react to the disturbances and the overshoot increases. If we set faster parameters for the controllers, the oscillations increase but the overshoot increases too. The only solution was to re-design the control loop.

We can also observe that pressure is proportional to temperature and the reaction is faster. This is due to the fact that when using steam to heat the milk, the pressure of the steam is directly linked to the heat exchange between steam and milk. The very interesting point is that pressure is reacting faster than temperature. It means that if we integrate the pressure in the control loop design, we can react to a disturbance before that disturbance impacts on the temperature and therefore limit the overshoot.

In addition we will use a model based controller (IMC). The main benefits of an IMC controller is that it embeds a mathematical model of the process which allows predicting the temperature behavior and reacting properly to disturbance.

This mathematical model is identified on site using dedicated software which calculates the mathematical model between actions on the valves and temperature reactions.

In order to validate our theory, we re-designed the control loop and integrated IMC controller and steam pressure control.Master Slave1


The temperature control loop (TIC) is managed by an Internal Model Controller (IMC) and the pressure control loop by a PID.

The milk temperature controller gives the pressure setpoint to the steam pressure controller. The overall process drawing will be the following:loop pic 2

After implementing this new control loop design we re-started the sterilization process in the same conditions as before.

The result was the following:Temp2

The red trend is the temperature measurement, the yellow is the valve setpoint, and the purple is the pressure measurement. We can observe a very significant improvement. First, 90% of the temperature setpoint is reached in less than 2 minutes, secondly there is no overshoot.

In the month following the test there were no stoppages as a result of the temperature setpoint going above or below the 5% margin of error.

Previously, the customer needed to reach a temperature setpoint of 140°C to meet the temperature of 138°C (UHT required setpoint temperature). With the new controller design that reacts very quickly to disturbances, and with limited overshoot, the customer has decided to reduce its temperature setpoint to 138°C

A month later, the controller re-design has delivered both energy savings, and an unexpected downtime reduction.

The results we have for the sterilization process are the same as the pasteurization process, as both are very similar processes.


Tags: , ,