Computer Control of Fermentation Processes

Daniel R. Omstead

• Publisher: CRC Press
• Year: 2020
• eBook ISBN: 9781000141757
• Edition: 1
• Topic: Computer Science
• Subtopic: Systems Architecture
• Index: Computer Science

The purpose of this volume is to describe the components, assembly, and implementation of computer-based process control systems. Presented in two sections, it illustrates how such systems have been used to monitor and control industrial fermentation processes as a means to improve our understanding of product biosynthesis. This book covers the fields of indirect parameter estimation and fermentation-specific control algorithms. It also includes chapters which describe system architecture and process application, process control, on-line liquid sampling and computer system architecture. This is an ideal source for anyone involved with biotechnology, bioengineering, microbial technology, chemical engineering, and computer control.

Chapter 1

TRADITIONAL SENSORS

David S. Flynn

TABLE OF CONTENTS

I. Introduction

II. Sensor Characteristics

A. Reliability

B. Accuracy

C. Precision

D. Response

E. Discrimination

F. Specificity

G. Sensitivity

H. Range

I. Maintainability

J. Availability

III. Sensor Design for Fermentation

A. Hygiene

B. Sterilization

C. Asepsis

D. Materials of Construction

E. Insertion and Removal

IV. pH

V. Dissolved Oxygen

VI. Temperature

VII. Pressure

VIII. Volume

IX. Flow of Gases

X. Flow of Liquids

XI. Agitation Rate

XII. Agitation Power

XIII. Foam

XIV. Redox

References

I. INTRODUCTION

Traditional sensors, by their nature, are ones which have been used in fermentation processes for decades to measure variables which were regarded either as important to control, or as providers of useful information. Unfortunately, perhaps as a result of their familiarity, their characteristics are assumed to be satisfactory^1,2 and hence, a cautionary word regarding their limitations is appropriate at this time.

II. SENSOR CHARACTERISTICS

The sensors used in fermentation share a set of characteristics common to those used in any process industry but require additional characteristics not often needed elsewhere. The shared characteristics include reliability, accuracy, precision, response, discrimination, specificity, sensitivity, range, maintainability, and availability. The additional characteristics include features such as materials of construction, means for insertion and removal, cleanability, sterilizability, and asepsis. Each of these characteristics are dealt with below.

A. RELIABILITY

The reliability of a sensor may be the most important characteristic since its credibility in the view of process operatives and managers largely determines its usefulness, regardless of its objective worth. Other important facets of reliability include physical robustness, failure rate, and failure mode.

The physical robustness of a sensor is largely a matter of a design which recognizes the operations which the sensor is expected to perform, the environmental conditions in which it is expected to operate, and the intentional or unintentional abuse to which it is subjected. As an example, a pH glass electrode is often fragile in itself, but when suitably shielded, it can withstand repeated steam sterilization, battering by particulate media in highly agitate tanks, and mishandling during insertion and removal.

Relatively little data appear to exist regarding the failure rate of traditional sensors,³ although Lees⁴ has published data collected from three chemical plants. His data suggest failure rates of 0.3 to 5.9 faults per year for instruments measuring temperature, pressure, flow, level, and pH, with pH the least reliable. When a sensor does fail, the way in which it fails, i.e., its failure mode, is sometimes vitally important, particularly if that sensor is part of a control loop. Provided that suitable constraints are built into the control loop, it may be better to suffer a sudden failure of a sensor than a slow or intermittent failure, since the latter modes may remain undetected for long periods. An example of a sudden failure is breakage of a thermocouple or failure of a weld on the diaphragm of a pressure transducer. Slow failures are associated with factors such as exhaustion of the electrolyte in a polarographic oxygen sensor or slow buildup of medium or microorganisms on a pH electrode.

Increases in reliability of a measurement can be achieved through replication of the sensors and conformity to good engineering practice. Good engineering practice includes proper specification of the sensor in the first instance, and installation and maintenance according to the manufacturer’s instructions. The cost of ownership of an improperly specified, installed, or maintained sensor is substantial when the effect of the misleading information it provides is taken into account.

B. ACCURACY

The accuracy of a sensor relates to the difference between the value produced by the sensor and some known or true value. Unfortunately, this simple statement hides a number of difficulties; for instance, over what time period is this comparison of values to be made, what indicated value is to be used (e.g., mean, mode, median), what is the true or known value?

There appears to be no generally accepted method of determining accuracy, and indeed some standards authorities such as the International Organisation of Legal Metrology (OIML) appear to avoid use of the term and speak only of “errors”. The practice adopted in the author’s laboratory is either to express the accuracy as the arithmetic difference between the mean indicated value and a “known” value (i.e., one determined independently of the sensor under scrutiny) over a time period which is relevant to the use to which that sensor is to be put (e.g., one fermentation, one shift), or to express the accuracy as the standard deviation of the differences between individual pairs of indicated and “known” values again over a relevant period, it is our experience that provided the definitions and specification are agreed upon at the start of an instrument trial, then the manufacturer or supplier is content to accept the user’s definition. However, when purchasing an instrument with no opportunity for a trial, there is sometimes no choice but to accept the suppliers statement of accuracy. The first problem with this is that the supplier may actually be quoting a value for precision (as defined below), and the second problem is that the value may be quoted as, for example, ± 1 %. Quoting the accuracy as a percentage is not sufficient in itself without expressing it as a percentage of the current reading or of the full scale, and without stating what proportion of the readings will be within this percentage. Discussion with the supplier may resolve these problems, but in the end it is for users to satisfy themselves regarding the performance of their instruments.

When an instrument is in routine use it is often not easy to check its accuracy, i.e., its calibration. For example, checking the calibration of a pH sensor installed in an aseptic fermenter is only possible either by removing the sensor for calibration and then replacing it without causing contamination, or by removing a representative sample from the fermenter and determining its pH before a significant change occurs in the pH of the sample. The former is possible by using the appropriate design of probe and holder although this is a cumbersome procedure; the latter is difficult to achieve, particularly in pressurized or deep fermenters where dissolved gases such as carbon dioxide will flash off as soon as the pressure is released.

So far in this section we have discussed only the accuracy of measurement. By contrast, in a control situation the accuracy may be expressed as the difference between the indicated value and the set point over a relevant period, and may be related to the characteristics of the controller and the actuator, as well as those of the sensor.

C. PRECISION

The precision of a measurement relates to the probability that repeated measurements will produce the same value. Once again, difficulties arise in interpretation since it is necessary to define, for instance, the method of measurement, the instrument used, the operator, and the laboratory, since they can clearly influence the result.

For a single sensor in a single situation, the practice in the author’s laboratory is to define the precision of measurement as the standard deviation of the indicated values over a relevant period. In a control situation, the definition of precision relates to the variability of the measured value about its mean value and hence will include the value for the precision of measurement.

One consequence of the adoption of definitions of accuracy and precision such as those given above for measurement and control situations is that it is possible to have examples of all combinations of high and low accuracy and high and low precision without contradiction of the definitions. Unfortunately, this can lead to confusion for the unwary without constant reference to the definitions.

D. RESPONSE

In a measurement situation, this relates to the time lag between the true and indicated value, and usually this time lag can be built up from a combination of transport delays and first-order responses. As an example, in the tubing technique for dissolved oxygen measurement which will be described later, there is a first-order delay caused by the diffusion of oxygen through the wall of the tubing followed by a transport delay caused by the time taken for the gas to travel from the coil of plastic tubing in the fermenter to the oxygen-measuring instrument, and then a further delay as the instrument itself responds to the change in oxygen content of the gas stream. For a simple situation with only a quasi-first-order response to consider, the response time is variously stated as the time constant (approximately 63% of the final value), or as the 90, 95, or 99% response time.

In a control situation, the response relates to the time lag between a change in the setpoint and the consequent change in the indicated value.

E. DISCRIMINATION

This is sometimes called the resolution of an instrument and relates to the smallest change in the indicated value. On an analog scale this value is largely a matter of opinion, but on a digital scale it is usually equated with a unit change in the least significant digit even though the latter rarely has any practical significance.

F. SPECIFICITY

This relates to the ability of the sensor to react only to the primary property being measured and be uninfluenced either by changes in other properties of the process or by the ambient conditions. It is necessary to distinguish between nonspecificity of the sensor itself and interference in the signal emanating from the sensor, although both are important in producing data which are not specific to the primary property.

Examples of nonspecificity in the sensor itself include the effect of temperature on pH measurement and the effect of pH on specific ion measurements. Interference in the signal from the sensor caused, for instance, by improper shielding or grounding is called electrical noise and is also capable of making an indicated value quite meaningless.

Specificity can be ensured by a combination of: (1) appropriate calibration (i.e., by taking into account the interfering process properties), (2) controlling the environmental conditions (e.g., temperature), (3) attention to good engineering practice (e.g., in installation), and (4) measuring the interfering influences and calculating the required property from the total data.

G. SENSITIVITY

The sensitivity of an instrument is variously described, but generally relates to the change in output of the sensor for a unit change in the measured variable. This relationship may not be constant across the whole range of the sensor and, hence, the output signal may need some processing, for example, by taking the square root or the antilog.

H. RANGE

This relates to the difference between the maximum and minimum values over which the instrument will operate. It is distinguished from the span of the instrument in that the span relates to that part of the range for which the instrument is set up for a particular application. As an example, the range of a resistance thermometer is – 200 to + 850°C, but it may be set up on an indicator or recorder with a span of, say, 0. to 100°C.

I. MAINTAINABILITY

This is a vital aspect of any sensor and relates to the time and effort required to recalibrate and repair it. In some applications, particularly with electronic equipment, the estimated time between failures and the estimated time to repair are quoted as MTBF (mean time between failures) and MTTR (mean time to repair).

J. AVAILABILITY

In a trivial sense availability is the ease with which the sensor can be acquired for use, but rather more importantly, it is the proportion of the working time of the process that the sensor is working properly. Clearly, in the latter sense a high availability is desirable and this is dependent upon both the reliability and the maintainability of the sensor.

III. SENSOR DESIGN FOR FERMENTATION

The sensor and actuator are the parts of the control loop which come into contact with the fermentation broth and, hence, need to be designed to facilitate hygiene, sterilization, asepsis, insertion, and removal.

A. HYGIENE

Most fermentations will be operated at least hygienically, that is, in a way which seeks to maintain a clean working environment. Therefore, it is important that the sensor is constructed to allow for easy cleaning and with a minimum of crevices which may harbor contamination...

Table of contents

• Cover
• Title Page
• Copyright Page
• Preface
• The Editor
• Contributors
• Table of Contents
• Chapter 1 Traditional Sensors
• Chapter 2 “Novel” Sensors for the Monitoring of Fermentation Processes
• Chapter 3 Automatic Fermentor Sampling and Stream Analysis
• Chapter 4 Indirect Parameter Estimation
• Chapter 5 Data Acquisition and Control Systems
• Chapter 6 Supervisory Computer Systems
• Chapter 7 Control Strategies for Fermentation Processes
• Chapter 8 Biological Modeling
• Chapter 9 Implementation of the Custom Built Fermentation Computer System at Eli Lilly and Company
• Chapter 10 The Application of Computer Control to Improve Fermentation Processes
• Chapter 11 Automating Batch Fermentations
• Chapter 12 Development of a Distributed Fermentor Process Control System
• Chapter 13 Computer Control of Recombinant Microbial Fermentations
• Chapter 14 On-Line Fermentor Sampling and HPLC Analysis
• Chapter 15 Upgrading an Existing Pilot Plant to Computer Monitoring and Control
• Index