Typical utility water systems are subject to considerable variation. Makeup
water characteristics can change over time. The abruptness and degree of change
depend on the source of the water. Water losses from a recirculating system,
changes in production rates, and chemical feed rates all introduce variation
into the system and thereby influence the ability to maintain proper control
of the system. Other variables inherent in utility water systems include:
||water characteristics (suspended solids, hardness, pH swings)
|| treatment product quality
These variables are considered and introduced during the applications and pilot
plant testing of new products for the treatment of various water systems. Pilot
plant simulation of actual operational variation is a challenging task. Every
industrial water system is unique, not only in the production operations it
supports and the sources of water it receives, but also in the degree of inherent
variation encountered due to the factors listed above. While a very sensitive
treatment program that must operate within a narrow control range may be suitable
for one system, another system requiring the same degree of protection may be
incapable of maintaining the required control. Consequently, inferior results
must be accepted unless the system is improved to support the sensitive program.
In operating systems, proper treatment of influent,
boiler, cooling, and effluent waters often requires
constant adjustment of the chemistry to meet the
requirements of rapidly changing system conditions.
A well designed program is essential to maintaining
proper control. The program should include
proper control limits and the ability to troubleshoot
problems that interfere with control of water
chemistry. Success in troubleshooting depends
on the knowledge, logic, and skills of the
troubleshooter. In order to improve operations it
is necessary to recognize the importance of
continuous improvement and to be familiar with
some tools and procedures necessary to support this effort.
Adequate and reliable data are essential if variation in a system is to he
measured and reduced. Specialized computer software can assist efforts to manage,
summarize, and use data effectively. Process data can be stored in a database
and retrieved and analyzed as needed in a variety of formats. Computers provide
nearly instantaneous access to many months or years of process data that would
require several filing cabinets if stored on paper log sheets. The computer
can he used to graph and analyze the data in a variety of formats, such as statistical
process control (SPC), trend analysis, and histograms. The operator is able
to troubleshoot the system based on these analyses without spending large amounts
of time manually researching and analyzing the data. In his classic hook Managerial
Breakthrough (McGraw Hill: New York, 1964, pp 1-14), Dr. J. M. Duran develops
the important distinction between quality control and quality improvement, and
describes the elements of effective problem solving in each case. These distinctions
and relationships are summarized in Figure
QUALITY CONTROL ZONE
Although the performance of a process varies from day to day, the average performance
and the range of variation are fairly constant over time. This level of performance
is inherent in the process and is provided for in the system design. The Quality
Control Zone in Figure 3-2 depicts
the accepted average and accepted range of variation in feedwater hardness.
This zone is often adopted as the standard of performance. Sometimes, performance
falls outside the accepted, or standard, range of variation in the Quality Control
Zone. This is depicted in Figure
3-2 by the sporadic spike. The goal of problem solving in the Quality Control
Zone is to reestablish performance within the standard. This involves the following
||detecting the change (sporadic spike)
|| identifying the cause of the change
||taking corrective action to restore the status quo
QUALITY IMPROVEMENT ZONE
Problem solving in the Quality Improvement Zone (also depicted in Figure
3-2) can have an even greater impact. The goal of quality improvement is
to reject the status quo as the standard and reach a level of performance never
before achieved. This level, the New Zone of Quality Control," represents
the achievement of lower costs and/or better performance. In this case, significantly
lower feedwater hardness decreases scaling potential and improves boiler reliability.
This step extends the scope of problem solving beyond the correction of obvious
problems. While it is important to "make the system work," it is often
more important to view the entire system to identify areas of potential improvement.
Some systems are poorly planned; others have not been updated to keep pace with
changing requirements and progressing technology. In either case, it is often
the system that causes control and operational problems not the people working
within the system.
Quality Improvement Tools
While a proper mindset must exist for continuous improvement, certain problem
solving procedures and tools can add structure and consistency to the effort.
The following quality improvement tools provide the means to summarize and present
meaningful data in a way that adds significance to the successful resolution
of chronic problems.
Flow Diagrams. A flow diagram provides a graphic presentation
of the steps required to produce a desired result. For example, this tool
may be used to clarify the procedures used to regenerate a softener or
the steps to be taken in the event of an upset in a cooling tower. Flow
diagrams are used in problem solving to give all parties a common understanding
of the overall process.
Brainstorming. In diagnosing a problem, new and useful
ideas can result when all of the people familiar with the process meet
to share their experiences and ideas. Possible causes are discussed and
possible solutions are presented and evaluated.
Cause-Effect Diagrams. An important first step in quality
improvement is the identification of the root causes of a problem. A cause-effect
diagram provides an effective way to organize and display the various ideas
of what those root causes might be. Figure 3-3
graphically presents possible causes for reduced demineralizer throughput.
Scatter Diagrams. A scatter diagram is useful in providing
a clear, graphic representation of the relationship between two variables. For
example, boiler feedwater iron levels might be plotted as a function of feedwater
pH to confirm or rule out a cause-effect relationship.
Pareto Analysis. Pareto analysis is a ranked comparison of
factors related to a quality problem, or a ranking of the cost of various problems.
It is an excellent graphic means of identifying and focusing on the vital few
factors or problems. Figure 3-4 represents an
analysis of the calculated cost of various problems interfering with the successful
management of a utility water system.
Meaningful Data Collection. Meaningful collection of data
and facts is fundamental to every quality improvement effort. Quality improvement
is an information intensive activity. In many cases, problems remain unsolved
for long periods of time due to a lack of relevant information. A good data
collection system must he carefully planned in order to provide the right information
with a minimum of effort and with minimal chance of error.
In order to plan for data collection, it is necessary to identify potential
sources of bias and develop procedures to address them:
||Exclusion bias. If a part of the process being investigated
has been left out, the result will be biased if the data is intended to
represent the entire process. For example, if data on attemperating water
purity is not included in an evaluation of a steam turbine fouling problem,
the cause could be missed.
||Interaction bias. The process of collecting the data
itself can affect the process being studied. For example, if an operator
knows that cooling tower treatment levels are being monitored by the central
laboratory, he may be more careful conducting his own tests.
||Perception bias. The attitudes and beliefs of the data
collectors can influence what they perceive and how they record it. If an
operator believes that swings in steam header pressure are his responsibility,
he may record that operation was normal at the time of boiler water carryover.
||Operational bias. Failure to follow the established procedures
is a common operational bias. For example, failure to cool a boiler water
sample to 25 °C (77 °F) often leads to an erroneous pH measurement.
Graphs and Charts. Pictorial representations of quantitative
data, such as line charts, pie charts, and bar graphs, can summarize large amounts
of data in a small area and communicate complex situations concisely and clearly.
Histograms. The pictorial nature of a histogram (a graphic
summation of variation in a set of data) reveals patterns that are difficult
to see in a simple table of numbers. Figure
3-5(a) is a histogram that shows the variation of inhibitor level in a cooling
water system. Each bar along the horizontal axis represents a specific range
of inhibitor concentration, in parts per million. The scale on the vertical
axis represents the number of occurrences within each range of concentration.
The shape of this particular histogram indicates a normal and predictable pattern
of distribution. There are no incidents of nonconformance outside of the specified
tolerance limits of 60-80 ppm, represented by the dotted lines.
In contrast, the patterns of variation depicted in Figure
3-5(b) and (c) represent problems, which must be corrected. The pattern
of distribution in Figure 3-5(b)
is relatively normal, but a few incidents of nonconformance occur outside of
the engineering limits, departing significantly from the otherwise normal distribution.
The cause of these occurrences must be investigated, and the process corrected
to a more predictable pattern. Figure
3-5(c) represents a normal and predictable pattern, but reveals several
occurrences that fall outside of the specified 60-80 ppm limits, indicating
that there is too much natural variation in the process.
Statistical Process Control. Statistical process control
(SPC) is the use of statistical methods to study, analyze, and control
the variation in any process. It is a vehicle through which one can extract
meaningful information about a process so that corrective action, where
necessary, can be implemented. While a histogram is a pictorial representation
of patterns of variation, SPC is used to quantify this variation and determine
mathematically whether the process is stable or unstable, predictable
or erratic. Figure 3-6
shows three SPC charts of the individual values of measurement used to
construct the histograms in Figure
3-5. In these cases, the data is plotted chronologically and used
interactively to determine whether a value falls outside of the statistical
With statistical process control, the actual historical data is used to calculate
the upper and lower statistical limits as a guideline for future operation.
Anything falling outside of the statistical limits is considered to be a special
cause of variation requiring immediate attention. Of course, if the common causes
of variation are excessive for either engineering or economic reasons, as is
the case in Figures 3-5(c)
and 3-6(c), improvement to
the process is necessary until the statistical limits are narrowed to the point