Since the initial use of computers in the 1960s for modelling fermentation processes (Yamashita and Mu-rao, 1978) and in process control for production of glutamic acid (Yamashita et al., 1969) and penicillin (Grayson, 1969), there have been numerous publications on computer applications in fermentation technology (Rolf and Lim, 1985; Bushell, 1988; Whiteside and Morgan, 1989; Fish et al., 1990). Initially, the use of large computers was restricted because of their cost but reductions in costs and the availability of cheaper small computers has widened interest in their possible applications. The availability of efficient small computers has led to their use for pilot plants and laboratory systems since the financial investment for the on-line computer amounts to a relatively insignificant part of the whole system.
Three distinct areas of computer function were recognized by Nyiri (1972):
1. Logging of process data. Data logging is performed by the data acquisition system which has both hardware and software components. There is an interface between the sensors and the computer. The software should include the com puter program for sequential scanning of the sensor signals and the procedure of data storage
2. Data analysis (Reduction of logged data). Data reduction is performed by the data-analysis su-tem, which is a computer program based on a series of selected mathematical equations. The analysed information may then be put on a print out, fed into a data bank or utilized for process control.
3. Process control. Process control is also performed using a computer program. Signals from the computer are fed to pumps, valves or switches via the interface. In addition the computer program may contain instructions to display deviccs or teletypes, to indicate alarms, etc.
At this point it is necessary to be aware that there are two distinct fundamental approaches to computer control of fermenters. The first is when the fermenter is under the direct control of the computer software. This is termed Direct Digital Control (DDC) and will be discussed in the next section. The second approach involves the use of independent controllers to manage all control functions of a fermenter and the computer communicates with the controller only to exchange information. This is termed Supervisory Set-Point Control (SSC) and will be discussed in more detail in the Process Control section.
It is possible to analyse data, compare it with model systems in a data store, and use control programs which will lead to process optimization. However, process optimization by this method is not a widely used procedure in the fermentation industries at present. It is important to be aware of these different applications, since this will influence the size and type of computer system which will be appropriate for the precise role that it is intended to perform, whether in a laboratory, a pilot plant, or manufacturing plant, or a combination of these three.
When a computer is linked to a fermenter to operate as a control and recording system, a number of factors must be considered to ensure that all the components interact and function satisfactorily for control and data logging. A DDC system will be used as an example to explain computer controlled addition of a liquid from a reservoir to a fermenter. A simple outline of the main components is given in Fig. 8.25. A sensor S in the fermenter produces a signal which may need to be amplified and conditioned in the correct analogue form. At this stage it is necessary to convert the signal to a digital form which can be subsequently transmitted to the computer. An interface is placed in the circuit at this point. This interface serves as the junction point for the inputs from the fermenter sensors to the computer and the output signals from the computer to the fermenter controls such as a pump T attached to an additive reservoir. Digital to analogue conversion is necessary between the interface and the pump T.
A sensor will generate a small voltage proportional to the parameter it is measuring. For example, a temperature probe might generate 1 V at 10°C and 5 V at 50°C. Unfortunately, the signal cannot be understood by the computer and must be converted by an analogue to digital converter (ADC) into a digital form.
The accuracy of an ADC will depend on the number of bits (the unit of binary information) it sends to the computer. An 8-bit converter will work in the range 0-255 and it is therefore able to divide a signal voltage into 256 steps.This will give a maximum accuracy of 100/256, which is approximately 0.4%. However, a 10-bit converter can give 1024 steps with an accuracy of 100/1024, which is approximately 0.1%. Therefore when a parameter is to be monitored very accurately a converter of the appropriate degree of accuracy will be required. The time taken for an ADC to convert voltage signals to a digital output will vary with accuracy, but improved accuracy leads to slower conversion and hence slower control responses. However, cycle times of about 1 second may be adequate in many fermentation systems. It is also important to ensure that the voltage ranges of the sensors are matched to the ADC input range. More detailed discussion is given by Whiteside and Morgan (1989).
A digital to analogue converter (DAC) converts a
digital signal from the computer into an electrical voltage which can be used to drive electrical equipment, e.g. a stirrer motor. Like the ADC, the accuracy of the DAC will be determined by whether it is 8-bit, 10-bit, 12-bit, etc., and will for example determine the size of steps in the control of rpm of a stirrer motor.
The small computer itself is dedicated solely to one or more fermenters. This computer is coupled to a real-time clock, which determines how frequently readings from the sensor(s) should be taken and possibly recorded. The other ancillary equipment linked directly to the computer might include a visual display unit, a data store, a teletype, a graphic display unit, a print out, alarms and a barometer.
The small computer is often connected to a large main frame computer for random access, not on a real-time scale, but for long-term data storage and retrieval and for complex data analysis which will not be utilized subsequently in real-time control.
It is also possible to develop programs so that on-line instruments can be checked regularly and recalibrated when necessary. Swartz and Cooney (1979) were able to routinely recalibrate a paramagnetic oxygen analyser and an infrared carbon dioxide analyser every 12 hours utilizing a program which connected a gas of known composition to the analysers and subsequently monitored the analyser outputs.
The simplest task for a computer is data logging. Parameters such as those listed in Table 8.1 can be measured by sensors which produce a signal which is compatible with the computer system.
Programs have been developed so that by reference to the real-time clock, the signals from the appropriate sensors will be scanned sequentially in a predetermined pattern and logged in a data store. Typically, this may be 2- to 60-second intervals, and the data is printed out on a visual display unit. In preliminary scanning cycles the values are compared with predefined limit values, and deviations from these values result in an error print out, or if more extreme then an alarm may be activated. In the final cycle of a sequence, say every 5 to 60 minutes, the program instructs that the sensor readings are permanently recorded on a print out or in a data store.
At the same time as on-line data is being recorded from sensors, analytical data for broth viscosity, microbial growth, substrate and precursor utilization and product formation, which have to be determined sepa rately may be logged into the data store for specific known times. Carleysmith (1989) has described a data handling system being used by Smith Kline Beecham in the United Kingdom.
Thus, it is now possible to record data continuously for a range of parameters from a number of fermenters simultaneously using minimal manpower, provided that the capital outlay is made for fermenters with suitable instrumentation coupled with adequate computer facilities.
Because a computer can undertake so many calculations very rapidly, it is possible to design programs to analyse fermentation data in a number of ways. A linked main-frame computer may be used for part of this analysis as well as the dedicated small computer.
A number of the monitoring systems were described as 'Gateway Sensors' by Aiba et al. (1973) and are given in Table 8.3. Gateway sensors are so called because the information they yield can be processed to give further information about the fermentation. More details of analysis of direct measurements, indirect measurements and estimated variables have been discussed by Zabriskie (1985) and Royce (1993).
The respiratory quotient of a culture may be calculated from the metered gas-flow rates and analyses for oxygen and carbon dioxide leaving a known volume of culture in the fermenter. This procedure was used to monitor growth of Candida utilis in a 250-dm3 fermenter, to follow or forecast events during operation (Nyiri et al., 1975).
If one defines the fraction of substrate which is converted to product then it is possible to write mass
Table 8.3. Gateway sensors (Aiba et al., 1973)
Information that may be determined from the sensor signal pH
Dissolved oxygen Oxygen in exit gas \ Gas-flow rate / Carbon dioxide in exit gas \ Gas-flow rate /
Oxygen-uptake rate 1
Carbon dioxide evolution rate , Sugar-level and feed rate " Carbon dioxide evolution rate ,
Acid product formation Oxygen-transferrate
Oxygen-uptake rate Carbon dioxide evolution rate Respiratory quotient Yield and cell density
, ,nccs for C H, O and N with the measurement of Î a few quantities (02, C02, NH3, etc.). All the lthcr quantities can be calculated, including biomass °id yield, if the biomass elemental composition is known (Chapter 2). This procedure was used for the analvsis of a bakers' yeast fermentation (Cooney et al, \q'ri). Biomass production can be regarded as a stoichiometric relationship in which substrate is converted, in the presence of oxygen and ammonia to biomass, carbon dioxide and water:
Carbon source-energy + oxygen + ammonium ->
cells + water + carbon dioxide. Thus, the equation can be written in the form: r/G. 11.(). + M)2 + CNH3 -» dCrHsO,N„
where a, b, c, d, e and / are moles of the respective reactants and products. C, H vO. is the molecular formula of the substrate where x, y and z are the specific carbon, hydrogen and oxygen atom numbers. Biomass is represented by CrHsO,N„ where r, s, t and u are the corresponding numbers of each element in the cell. This technique was developed to use with bakers' yeast fermentations (Cooney et al., 1977; Wang et al., 1977).
Arminger and Moran (1979) recognized three levels of process control that might be incorporated into a system. Each higher level involves more complex programs and needs a greater overall understanding of the
Was this article helpful?