Problems with Computer Control

The use of computers and microprocessors (also known as programmable electronic systems [PES]) in process control continues to grow. They have brought about many improvements but have also been responsible for some failures. If we can learn from these failures, we may be able to prevent them from happening again. A number of them are therefore described below. Although PES is the most precise descnption of the equipment used, I refer to it as a computer, as this is the term usually used by the nonexpert. 

---

Kletz, Trevor. Human problems with computer control. Plant/Operations Progress 1 (4), October... 

T.Kletz, Human problems with computer control i/azardPrevention, March/April, 24, 1983. 

Besides the above-mentioned problems with step control, there are also other computational aspects which tend to make the straightforward NR problematic for many problem types. The true NR method requires calculation of the full second derivative matrix, which must be stored and inverted (diagonalized). For some function types a calculation of the Hessian is computationally demanding. For other cases, the number of variables is so large that manipulating a matrix the size of the number of variables squared is impossible. Let us address some solutions to these problems. 

However, even this approach becomes inefficient if the degrees of freedom and the reduced Hessian matrix become large. Consider a problem with NU control variables per element and NE finite elements. Note that even after decomposition, a Newton method applied to a dense, reduced system with NU X NE degrees of freedom requires computational effort on the order of (NU X NE) ... 

The control computer/DCS system consists of controllers, A/D and D/A converters, and the signal conditioifing hardware and software, i.e., filtering and validation. Each of these components requires separate evaluation. Table 15.5 lists possible problems with the controller/DCS system. One way to initially check controller tuning is to place the control loop in manual (open the control loop) and observe whether the controlled variable lines out to a steady-state or near steady-state value. Comparing the open-loop and closed-loop performance indicates whether the controller is upsetting the process. If not, disturbances to the control loop in question are the primary source of the upsets. 

Automated analysis of the data according to this equation, such as by computing the linear regression of In vs. t, can be used to extract R and Q (see Problem 15.12). Once Ryi is known, one can adjust the value of/in the positive feedback circuitry to approach unity in a systematic way while testing for early indications of potentiostatic instability. All of this can be done automatically with computer-controlled instruments. 

Already in the mid-1960s, there was rich potential of applying such experiments to the determination of concentrations but even more to the elucidation of reaction mechanisms and kinetics coupled to electron transfer at an electrode was recognized. Today the resulting Knear sweep or cyclic voltammetries are employed as simple, flexible routine techniques in particular as sophisticated means to solve chemical and mechanistic problems. The combination with computer control, ultramicroelectrodes, and digital simulation has further contributed to their success. 

Gerchak, Y. and M. Parian 1987. A single period inventory problem with partially controllable demand. Computers and Operations Research, Vol.14, No.l, 1-9. 

Machinery should be built in such a way as to minimise the exposure of workers to the cutting fluids. Manual feeding of workpieces, oil contact when machines are shut off and hands are placed into the machines, and the need to hold workpieces under the flow of oil when it is being machined are machine-design problems which will increase the risk of development of dermatitis. Increasing automation with computer-controlled machines which are enclosed to prevent oil contact will reduce the incidence of dermatitis. 

The most spectacular successes of X-ray methods, however, are in molecular and crystal structure analysis. Examples are the structures of insulin [130], hemoglobin (131], and vitamin B 2 [132]. Today, single-crystal X-ray structure analyses of relatively complex compounds (up to ca. 40-200 nonhydrogen atoms) are performed with computer-controlled diffractometers whose computers can be used simultaneously to solve the phase and structure problem within a few days or just a few hours.. The method has gained parity in investigation time with spectroscopic methods... 

For optimization problems with two control variables there can be three different possibilities at any given point of time (i) only Fj is singular, (ii) only F2 is singular, and (iii) both Fi and F2 are singular. If only Fi is singular and F2 is either at maximum or minimum value, the F can be computed simply by setting dSi/dt = 0 in Eq. 2. If we include the correction term, this results in ... 

The advancement of modern electronics coupled with computer control has now concentrated the problems of DMTA measurements completely at the mechanics end. Selection of optimum geometry for the materials to be measured is thus vitally important. Polymers sag under their own weight and flow away from clamp pressure in the Tg region and above. The 1000-fold stiffness change for amorphous systems through Tg makes the experimental side of DMTA quite difficult and the best instruments have to be judged on... 

Note The creation of multiple DGs is not practical for large problems with complex control flow, due to longer computation times and increased memory requirements. 

One of the most severe problems with using control bits is that it is not possible to make loops in a simple way. In particular, it is not possible to let exception criteria depend on the state of the computer. However, this is essential for universal computing. Thus, it is necessary to let the path of the computation depend on the state of the computer rather than on the state of the control bits. This is possible Control bits are not necessary to perform local deterministic computation as will now be shown. 

Derivative methods work well only when sufficient data are recorded during the sharp rise in plT occurring near the equivalence point. This is usually not a problem when the titration is conducted with an automatic titrator, particularly when operated under computer control. Manual titrations, however, often contain only a few data points in the equivalence point region, due to the limited range of volumes over which the transition in plT occurs. Manual titrations are, however, information-rich during the more gently rising portions of the titration curve before and after the equivalence point. 

Load sharing or selective load shedding is of interest to many users of hot gas expanders. A particularly successful European FCC application is illustrated in Figure 6-43. The addition of an expander-generator set to the FCC unit at a major refinery presented a challenge because a trip of the expander could upset the process. The company that is the subject of this application case study, GHH Borsig, solved this problem with the installation of a computerized control system and through computer simulation of trips. 

It is always risky to keep only one copy of a document. If computer generated, you can easily make another copy provided you always save it, but if manually generated, its loss can be very costly. It is therefore prudent to produce additional copies of critical records as an insurance against inadvertent loss. These insurance copies should be stored in a remote location under the control of the same authority that controls the original records. Insurance copies of computer disks should also be kept in case of problems with the hard disk or file server, if you use one. 

The complete packing process is computer monitored and computer controlled. Potential problems in column packing can be seen directly and the affected column is removed from the production cycle. Pig. 9.11 shows a graphic representation of a column packing process by monitoring the packing pressure with time. 

In many cases, problems cannot be overcome by biological means. This is especially true for those related to inhibition by substrate or product. There may, however, be technical solutions to these problems. Nowadays, complicated feed strategies with different substrates can be achieved through the use of flow injection analysis, on-line sensors, mass flow meters and sophisticated computer control. Such control coupled to a fed-batch mode of operation (Figure 2.5) can often eleviate problems caused by substrate inhibition. For some processes, continuous product removal can avoid the problems associated with product inhibition the various options include ... 

---