Modeling and Control

When DeltaV Predict (embedded model predictive control, MPC) was introduced in 2001, I developed some of the MPC lecture material and process simulations that are used in our advanced control education class, 7201. Also, I helped teach the first three of four classes. I always enjoy teaching in these types of classes since of the customers who attend this training often represent a variety of industries. In one of one of the classes, a student showed a particularly strong interest in MPC and discussed several areas of his process that would benefit from this technology. It turned out that the student was the head of instrumentation at a Pharmaceutical plant in Puerto Rico.

A few months after this class, our service department received an order from this pharmaceutical customer to install MPC in one process area. Since DeltaV Predict was new to our service organization, the advanced control team provided some support in the implementation and startup of this MPC application. The savings from this one application were substantial (in the seven digits) since the area was a bottleneck in plant production. We later shot a video at the plant site. In this video the customer discusses the installation and the benefits they realized for this installation of MPC technology.

Making a control change, such as the installation of MPC, in a pharmaceutical installation was at that time complicated by the fact that the target process was an existing validated installation. Fortunately, our service department had done the original installation and thus was very familiar with the regulator submissions needed to make this change in the process control. However, the paper work required to make this improvement using proven embedded technology still took a consider amount of work and time. Because of this paperwork, many manufacturing procedures in the pharmaceutical industry are often treated as being unchangeable.

Some time after this MPC installation was completed, the Process Analytical Technology, PAT, initiative was established by the Federal Drug Administration (FDA) to encourage the voluntary development and implementation of innovative pharmaceutical development, manufacturing, and quality assurance. As part of this framework, an innovative approach has been developed for helping the pharmaceutical industry address regulatory issues and questions. This new approach is discussed in Guidance for Industry, PAT – A Framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance. Through this initiative, the pharmaceutical industry has been given an opportunity to more easily innovate and improve operations through the application of proven measurement and control techniques.

The control design and commissioning of a continuous process often focuses on the requirements associated with operating conditions found at normal plant throughput. Thus, during the startup of the process, many controls are left in manual and it is up to the operator to get the process to the point where the controls can be placed in automatic. From a practical standpoint, it makes sense that the design address normal operating conditions since this is where the process will operate most of the time. Thus, the problem of non-linear installed characteristics of final control elements, limited rangeability of measurement elements, etc are often ignored. The startup of a batch process has many of the same challenges associated with startup of a continuous process. However, since the startup phase of a batch process often represents a significant portion of the total batch cycle time, it is important to automate this phase of the operation. From this perspective, the design and commissioning of batch process controls can often be much more demanding than for a continuous process. Addressing the challenges associated with the wide range of operation during the startup phase of a batch process can pay benefits in terms of reduced cycle time. For processes that are not limited by market demand then any reduction in cycle time directly translates to increase production and associated profit. In some cases, manufactures have added features to the PID that may be used to address some of the requirements associated with the startup of a batch process.

It is common for one or two loops associated with a batch operation to exhibit a slow response. For example, the multiple lags associated with a critical temperature measurement may introduce the equivalent of a signification delay in the process response. Thus, this slow response may require that a long reset time be used in the temperature control. This slow tuning may be appropriate once the unit has reached the design processing conditions. However, during the batch startup this tuning may limit how quickly the unit may be brought up to design processing conditions. The time required for the startup phase of a batch may often be reduced using batch logic at the start of the batch to switch the control to manual and then to position the temperature controller output to a value required for normal operations. Once the impact of the change in the heating valve starts to be reflected in the batch temperature, then batch logic can be used to switch the control to Automatic when the temperature reaches a certain point. However, such a procedure is often complicated by the fact that a different valve position and point to switch the control to automatic may depend on the initial process conditions e.g. feed temperature or the product grade.

To assist a batch process in reaching normal operating condition in minimum time without the need for special batch logic, the DeltaV PID has two added parameters which were named anti-reset windup high and low limit, ARW_HI_LIM and ARW_LO_LIM. The PID is designed to address reset windup independent of these limits. Thus, the real purpose of these ARW parameters is to allow a batch process to quickly be brought on-line. For batch applications, these limits may be configured inside the controller output limits. When this is done, the controller reset time is reduced by a factor of 16X if the controller output is outside the ARW limit and the sign of the control action is driving the output toward the ARW limit. Thus, the control may be switch to Automatic mode at the start of a batch and the control output changes due to reset action will occur 16 times faster until the output reached the ARW limit and then the control will revert to normal reset action. This control action is illustrated in the following.

Response Using ARW limits

For batch applications that require this type of action during startup, the ARW limits should be set to values that are outside the range that will be required during normal batch operations. The ARW limit value may be adjusted to reach setpoint in minimum time without overshoot.

In some continuous applications, the process may be operating at saturation i.e. full open or closed valve under certain operating conditions. When the operating conditions abruptly change, then it may be necessary for the control to quickly position the valve of a slow action loop to avoid the associated control parameter from overshooting. In such applications, the ARW limits may be set to improve recovery from these process saturation conditions and get to setpoint without the use of special logic.