A Programmable Logic Controller (PLC) was used for data acquisition and to control the primary actuators. Operator and control calculation interaction were carried out via an IBM-compatible personal computer, linked to the PLC. Project-specific software was developed for the personal computer with a graphic interface to handle the control systems in either automatic or manual mode.
The primary objective of the control system was to regulate the average temperature of the solid bed at a fixed value and to control the bed water content (according to a varying set point). The secondary objectives were to minimize temperature gradients within the bed and also to prevent the bed from becoming overly compact.
Control of the bed temperature was based on evaporative cooling, by manipulating the relative humidity of the inlet air and maintaining its temperature at a low value. During the period of high heat generation, in order to avoid bed overheating, it was necessary to intervene manually to manipulate the inlet air flow rate and temperature and to initiate agitation events. The average bed temperature was fed into a digital PID control algorithm to drive the set point of the inlet air relative humidity. This in turn was controlled with an on/off algorithm with a dead band and hysteresis that commanded a solenoid valve adding steam. The inlet air temperature was further controlled with another on/off algorithm with a dead band and hysteresis that manipulated the heater or cooler, according to process needs.
The water content of the bed was controlled through the periodic addition of fresh water. The reference trajectory of the water content was computed in the laboratory based on water activity studies. A solid sample was taken each hour from the bioreactor for an off-line measurement of the water content. The amount of water to be added was determined by the operator using an approximate water balance and his experience. The bed was agitated upon each addition of water.
In order to keep the bed as homogeneous as possible and to avoid excessive inter-particle aerial growth, which would reduce porosity, a periodic agitation policy was established. The degree of homogeneity was defined by the temperature gradient inside the bed, while inter-particle aerial growth was estimated from the pressure drop through the bed. This involved a semi-automatic loop that employed heuristic logic. The operator could establish the agitation speed, its duration and, in the case of the 200-kg bioreactor, the path of the agitation (left or right).
The control strategy enabled efficient operation of the pilot SSF bioreactors. When the 50-kg bioreactor was run manually, it required the permanent attention of at least two operators. The automated control system required only one operator, even at the 200-kg scale, who intervened relatively little in the process. The system even allowed bioreactor operation without direct supervision at certain times (Fernández 2001). Figure 28.2 shows the performance of the average bed temperature control loop between hours 10 and 40 of a fermentation run in the 200-kg bioreactor. The control system performed reasonably well, since the average bed temperature deviated no more than 4°C from the set point, although most of the time the deviations were smaller than 1°C. However, to achieve this performance, the inlet air temperature had to be changed manually. Moreover, the differences between maximum and minimum temperatures within the solid bed were considerably high (5°C average difference and 15°C maximum differences).
On the other hand, water content control did not perform so well, with deviations of more than 25% with respect to the set point. This was due to the several limitations that this control loop presented, such as manual control, the lack of an on-line sensor and the fact that water had to be added while the bioreactor was being agitated such that control actions could only be taken infrequently. In addition, water content measurements were noisy due to bed heterogeneity.
However, it is noteworthy that the control strategy was successfully scaled up from the 50-kg bioreactor to the 200-kg one with only minor adjustments.
The operation of the bed temperature control loop can be simplified if a model predictive control algorithm is used. When this kind of control was applied in the 200-kg bioreactor, within the same time-span shown in Fig. 28.2, better overall performance was achieved. Temperature differences within the bed and high temperature peaks were reduced when compared with standard control (Fig. 28.4). In addition, the loop was fully automatic therefore no manual operation of the inlet air temperature was necessary.
It should be noted that even better performance could be achieved by tuning the algorithm; however this was not done with the PUC bioreactors since it would have required several fermentation runs, each of which is long and expensive. Despite this, it is probably correct to say that, to attain good performance in industrial-scale SSF bioreactors (2 to 3 tons or more), it is necessary to apply model predictive control in the bed temperature control loop.
10 15 20 25 30
10 15 20 25 30
Fig. 28.2. Bed temperature control during a fermentation run with the 200-kg bioreactor. Key: (--) Bed temperature set point; (—) maximum bed temperature; (•), average bed temperature; (- - -) minimum bed temperature
Was this article helpful?