Batch Optimization Perspective Tips
The highest value added products use batch operations. Batches can take days to complete and be worth millions of dollars. In many cases bad batches cannot be fixed downstream. Bad batches must be avoided. There are many techniques for making batches more repeatable and faster by better monitoring and control. Essential are an appreciation of automation and an understanding of the non-self-regulation inherent in a batch process and the implications in terms of control strategy and controller tuning.
Pharmaceutical and specialty chemical manufacturing tend to use batch unit operations because these mimic lab bench top operations in product research and development (R&D) offering the shortest time to market. The pharmaceutical industry is starting to change goals. Big driver for batch is minimizing cross contamination risk and bracketing of out of spec material so it can be quarantined. Additionally, the batch can be held until the product meets specifications allowing for time for conversions and separations to complete and for operations to make corrections. Small volume high value added products with patent protection benefit the most from batch operations. The extreme example is biological pharmaceuticals where each batch is worth millions of dollars, total yearly production may be less than a 100 kg and the clock is ticking on patent protection. The other major factor preventing continuous bioprocesses is the buildup of inhibiters and contaminants, viruses, harmful bacteria, and non-viable, dead and mutated cells.
While continuous operations can provide greater capacity, some of the feed or incompletely processed material in well mixed volumes is being discharged reducing yield and necessitating in many cases recovery and recycle. The fact there is no liquid discharge flow in batch operations until the end of the batch leads to a lack of self-regulation that is the key to the distinctive features of batch dynamics and the unique control system requirements. While some reactions can be self-limiting in that reactant consumption is functionally similar to a discharge flow, unbalances (e.g. excess reactant per stoichiometry) can cause a buildup of reactant. Consequently, batch composition, pH, and temperature response join the gas pressure response as integrating or runaway processes. The continual accumulation feeds and products by separation and conversion results in process conditions changing throughout the batch. There is no steady state in the conventional sense. The response may be in one direction only (e.g. single ended). The dynamics are nonlinear. Since model predictive control (MPC) thrives on steady state linear processes and requires a response in both directions, a creative translation of controlled variables is needed. One such possibility is the use of the slope of batch concentration, pH, or temperature profile as a controlled variable. The profile slope has a steady state, is more linear, and can decrease besides increase in value.
Valve position control (VPC) and override control are commonly used for optimization of batch cycle time. These advanced regulatory control systems are less dependent upon a process model and can be readily configured using existing process variables to achieve simple optimizations such as the maximization of feed rate for fed-batch operations.
In pure batch operations, the feeds are sequenced on and off based on total charges and batch logic. Phases of pressurization, heating, and cooling are sequenced as well. Temperature, pH, and pressure control loops are cycled in and out of service.
In fed-batch operations, the feed rates are typically manipulated by flow loops. Often flow ratio control is used. An analyzer can be used to provide a higher level of control to correct the ratio to maintain the stoichiometry. The ratio of utility to feed flow may also be used to provide a desired vaporization, heating or cooling rate. Fed-batch operations create the opportunity for more feedforward and feedback control and optimization opportunities. Some call fed-batch semi-continuous because there is a throttled rather than a sequenced flow rate. Since there is no liquid discharge flow, the process response is still integrating or runaway. In fed-batch operation, the variability in the composition profile is transferred to variability in the manipulated feeds. Many process engineers are reluctant to turn over the transfer of variability to a control loop. Most want to fix both the feed rates and the composition not realizing that variability doesn’t disappear but is transferred from the controlled variable (composition profile) to the manipulated variable (feed flow). Some process engineers try to duplicate some of the benefit of fed-batch operation by scheduling the flow rate and timing of a sequenced feed at opportune points.
Unlike continuous processes, setpoint response rather than load response is most important. Often the overshoot is more critical than rise time (time to reach setpoint). Some applications might require increasing the batch temperature to help dissolving additives; if the product of the batch needs to be fed at lower temperatures to the next process stage, an overshoot in batch temperature represents an unnecessary and costly waste of energy. For bioreactors with mammalian cell cultures the overshoot typically must be less than 0.05 degree Centigrade and 0.05 pH in the temperature and pH loops, respectively, for a shift in the setpoint to optimize product formation. The time to reach setpoint in these bioreactors is a small fraction of the total batch cycle time of 10 days or more. Also, achieving specified quality is more important than reducing batch cycle time. In many cases, the batch time is fixed and conservatively set for biologic products. For these applications where rise time is not important, a setpoint filter equal to the reset time or a structure of PD on PV and I on error is an effective solution for eliminating overshoot, provided a tuning method such as lambda tuning for integrating processes is used so that correct balance between integral mode and proportional mode is maintained. The balance is expressed by the product of the PID gain and reset time being greater than twice the inverse of the open loop integrating process gain.
For more mature and higher volume products, batch yield and capacity tends to become more important. To reduce batch cycle time, sequential operations can be replaced with simultaneous operations. Time intervals between actions can be reduced and actions can be intelligently automated, such as the automatic detection of end points and initiation of the next phase. Valve positions of both on-off and throttle valves can be maximized. Setpoint rise time can be minimized by intelligent scheduling of PID outputs. Yield can be increased by optimization of setpoints and feed rates based on raw material analysis and feedback correction by composition profile control or analysis of batch end points.