ANALYTICAL CHEMISTRY, VOL. 51, NO. 11, SEPTEMBER 1979
1739
Characterizing Chemical Systems with On-Line Computers and Graphics Jack W. Frazer,” Lester P. Rigdon, Hal R. Brand, and Charles L. Pomernacki Lawrence Livermore Laboratory, University of California, Livermore, California
Incorporating computers and graphics on-line to chemical experiments and processes opens up new opportunities for the study and control of complex systems. Systems having many Variables can be characterized even when the variable interactions are nonlinear, and the system cannot a priori be represented by numerical methods and models. That is, large sets of accurate data can be rapidly acquired, then modeling and graphic techniques can be used to obtain partial Interpretation plus design of further experimentation. The experimenter can thus comparatively quickly iterate between experimentation and modeling to obtain a final solution. We have designed and characterized a versatile computer-controlled apparatus for chemical research, which incorporates on-line instrumentation and graphics. I t can be used to determine the mechanism of enzyme-induced reactions or to optimize analyticai methods. The apparatus can also be operated as a pilot plant to deslgn control strategies. On-line graphics were used to display conventional plots used by biochemists and three-dimensional response-surface plots.
We have been working to develop advanced computercontrolled instrumentation and graphic techniques to use on-line in complex experiments and chemical processing. One goal is to be able to rapidly characterize nonlinear chemical systems having many variables. A second goal is to use the characterization data and models to develop control algorithms. The coefficients of these control algorithms are to have a known relationship to the chemistry being controlled. We have chosen a study of enzyme analysis to demonstrate our system. Several approaches to the automation of enzyme analyses have been published (1-4). The instrumentation and techniques described here are also applicable to the study of enzyme mechanisms as well as analytical methods. For most chemical experimentation, manual control of instrumentation followed by off-line manipulation of the resultant (small) data base is no longer adequate. A better understanding of the kinetics and thermodynamics of chemical processes can be obtained by controlling experiments with inexpensive on-line computers. On-line instrumentation opens up new opportunities for the study and control of “medium” number systems (5) (one having organized complexity and frequent discrepancies with any existing theory). That is, we can now more adequately study systems containing many parameters even when the parameter interactions are nonlinear and the systems cannot a priori be represented by analytical methods. The opportunities are unbounded for the analytical chemist. His expertise in the development of chemical instrumentation and methods of analyses provides him the background required to solve many experimental and process-control problems. Moreover, instruments are now being designed for use on-line to experimentation and processes. Thus, we are entering a period in which computers will enable the incorporation of analytical instrumentation into the experiment or process. 0003-2700/79/035 1-1 739$01 .OO/O
94550
Chemical research will take increasing advantage of these developments. Among the many benefits to be derived are: improved analytical accuracy, experimentation that otherwise would be difficult to perform, unattended operation, on-line graphics, and enough data to numerically model the processes being studied. While opening up many new possibilities, this more advanced approach to experimentation also raises new problems such as characterization of the apparatus and development of new experimental design strategies. T o be able to perform such experimentation one needs to properly specify and design (6-8) the apparatus to take full advantage of the computer(s) and instrumentation. Automated systems can often rapidly generate the large volumes of data required to characterize complex systems. Interpretation of these vast quantities of data by conventional techniques is very time consuming. However, the use of graphics on-line to complex experimentation can greatly aid data reduction, interpretation, and development of numerical models. As an example, the experimental data can be plotted as a function of any required number of sets of two or three variables. Also, numerical models of the system together with dynamic graphic representations can be developed and used to improve the scientist’s understanding of the chemical and physical processes. Research in these areas will eventually result in the development of techniques that will support the real-time analysis of the dynamics of chemical systems. Consider the development of analytical methods, where it is sometimes necessary to investigate five parameters and their relationships. If some of the relationships are nonlinear, a reasonable factorial design might require 6 steady-state measurements per parameter or 7776 (S5) measurements. Given a completely automated apparatus capable of obtaining 1 datum in only 2 min it would still require over 259 h of uninterrupted effort to acquire the data (this allows no time for quality control or instrument calibration). Modeling and analysis of such a data set is very difficult and time consuming. To reduce the time and difficulty of interpreting complex data sets, more research effort must be directed toward coupling numerical modeling with self-modifying control algorithms; this will result in the acquisition of the minimum data necessary to characterize the process to a specified accuracy. To support research of the type discussed above, a versatile dynamic apparatus for chemical research has been designed to incorporate on-line instrumentation, on-line graphics, and computer control of the experimentation. However, accurate interpretation of the data from this apparatus requires that the time-response characteristics of the system be carefully modeled. Therefore, the apparatus and instrumentation were f i t characterized in terms of their dynamic transfer functions. The results of these characterizations were then incorporated in experimental designs for the investigation and characterization of rapid-equilibrium and steady-state enzymecatalyzed reactions. The apparatus and its characterization, along with the more conventional enzyme kinetics plots and three-dimensional plots of response surfaces, will be discussed in this paper. Subsequent papers will describe the investigation of exper0 1979 American Chemical Society
1740
ANALYTICAL CHEMISTRY, VOL. 51, NO. 11, SEPTEMBER 1979 Temperature-control led b a t h i
Reagents-
0 0 0 0 0 0
0 0
1
0 000 000
Spectrophotometer
loo0
lo00
f valves
I
Reference side
Pumps
I
I
I
motors drivers
Keyboard terminal
I
II
II
Delay 1 oop valves
Digital thermometer
Keyboard terminal
imental design strategies, the development of new graphic techniques as aids in model development and identification of experimental errors, reaction characterizations using threeand four-dimensional models, the development of control strategies, and the development and optimization of analytical methods. Although an understanding of enzyme mechanisms is of great importance, these studies were chosen principally as a vehicle for developing computer-controlled experimental techniques required t o characterize and model complex nonlinear systems and process control algorithms based on t h e chemical characterization.
APPARATUS Figure 1 is a block diagram showing the entire processing and control system. The system consists of: (1)reagents, (2) three-way valves for system cleanup, (3) a bank of pumps, (4) delay coils to allow solutions to reach temperature equilibrium before mixing, ( 5 ) micro (160 ILL)magnetic mixers, (6) delay loops and valves to allow easy extension of the chemical reaction time, and ( 7 ) continuous flow spectrophotometer systems to allow spectrophotometric measurements of the product(s). The flow system from the pumps to the detectors is suhmerged in R thermostated water bath. The system shown has two sections that can be operated independently or in parallel as an experimental aid to test the effect of important variables on product yield. In the parallel mode the separate sides are operated identically except one (or more) variable on one side of the system is operated with a gain or offset or combination of the two. Thus, the effect of such gains or offsets on product yield can quickly he assessed. P u m p s and P u m p Tubing. Sixteen peristaltic pumps (Cole-Parmer Instrument Co. Model No. 7013) are mounted on individually controlled stepping motors (Superior Electric Type SS250-1002), which can he driven from 0 to 60 rpm in increments of 1/200 of a revolution per step. The rate of each pump can be ramped up or down through the range over a specified time period,
or maintained at a fixed rate by a computer control algorithm. We use nominally 0.0315-in. i.d. silicone pump tubing (ColeParmer No. 6411-41), which will deliver about 2.5 X mL per step or 3 mL per min at the maximum pumping rate. The actual delivery rate of each pump is calibrated by weighing the amount of water pumped at a specified rate for a specified time. Mixers. Two 160-pL-volumemagnetic mixers (Bar-Fran Corp., San Diego, Calif.) with nine ports each are used to mix incoming solutions for the reference and reaction sides of the flow system. The time allowed for reaction is calculated from the volume of the flow line (including any delay loops selected) from the mixer to the detector and the total pumping rate. The mixers can he turned on or off by the computer. Delay Loops and Valves. Four delay loops and valves (Altex Scientific, Inc. Berkeley, Calif.) are used on the reference and reaction sides of the flow system to control the time allowed for reaction. The valves are controlled by the computer and may be selected in any combination. As shown in Figure 1. the solution may go straight through, or it may be directed through any combination of loops. The delay loops have volumes of 2, 4, 8, and 16 mL and the straight-through path has a volume of 1 mL. Thus, the volume of the delay lines may be selected in increments of 2 mL from 1 to 31 mL. Detectors. Two dual-beam spectrophotometers (Schoeffel Instrument Corp. Model SF 770) with a wavelength range of 200 to 630 nm are used to monitor the output product(s). The spectrophotometers are equipped with 1-mm bore, 10-mm path-length quartz optical flow cells and deuterium or tungsten light sources. The monochromators are computer controlled, and the signals from both the reference and sample sides are read by the computer via an A/D converter. Flush Out Valve and Flow Tubing. Each reagent is aspirated from its reservoir through a three-way valve, which may be switched by the computer to aspirate reagents or to flush out the pump tubing and flow lines with water or a cleaning solution. The reagents flow through the pumps into a delay coil so that they will reach bath temperature before reaching the mixers. The
ANALYTICAL CHEMISTRY, VOL. 51. NO. 11, SEPTEMBER 1979
flow tubing is nominally 0.031-in. i.d. Teflon, while the delay loops are 0.060-in. i.d. Teflon. Operating Characteristics. The rates of rotation of the pumps are individually controlled by the PDP 11/45 computer and each pump has a software velocity meter and accelerometer. The rate of delivery is limited at the lower rate by the required blending accuracy and at the higher rate by a maximum stepping motor speed of 60 rpm. For our system we find by numerical modeling and experimentation that blending ratios equal to or less than 1O:l must be maintained if the delivered square wave of solution is to be blended with the other streams to produce