You are on page 1of 16

Food and Pharmaceutical

Division Winter2013

In this Issue
1 Directors Message
2 Networking at ISA
2 Anti-Patterns
3 Alarm management technical report published
3 Alarm Management for Food / Pharmaceutical
Processes
7 Division Meeting at FLM Director
3 ISA Automation Week 2012 Alex Habib, PE
8 Models and Support for Alarm Operation;
Batch vs. Continuous
14 2012 Conference Survey Editor
15 New Members
Andre Michel, PE
16 Division Officers


A message from our Division
Process Analytical Technology (PAT)
Virtualization
Director Alarm Management
Batch Process Controls
Dear members of the Food and Pharmaceutical Industry
Cyber Security
Division (FPID);
Lot Tracking
On behalf of the Food & Pharmaceutical Division, I am Food and Pharma Safety & Security
pleased to confirm that our FPID division board decided to Bio-Pharmaceutical Automation & Process
hold a Food & Pharma symposium in the spring of 2014 at Controls
ISAs Head Quarter at RTP. Control System Validation
Food & Pharma Instrumentation Systems
The decision was based on the following main reasons: (hardware & Software)
1- North Carolina is one of the main centers for the
manufacturing and research facilities of food,
pharma and bio-tech industries in the USA Sincerely
2- ISAs Head quarter provides an excellent meeting Alex Habib, PE
facilities and good support by our staff. FPID Division Director
3- Several ISA divisions had excellent participation 732-679-1887
by operating companies and vendors for example: Alex-habib@msn.com
the Analytical Division, Power Industry and
Water/Waste Treatment, etc.
4- ISA has excellent choices of hotels in the RTP area
at reasonable cost.
5- Our Tar Heel section has a good relationship with
other pharmaceutical societies such as ISPE.
In order to make this Symposium a success; we need your
help to find qualified speakers and potential sponsors.

Please contact me at your earliest convenience if you are


interested or know of a speaker to cover such items like:
Networking at ISA Automation share your thoughts, or if you could not make it and want
to offer your insights on planning next years event, drop
Week 2012: Real Conversations us an email at info@isa.org. We are always interested in your
opinions and feedback.
Start here
By Robert E. Lindeman, 2012 ISA President Anti-Patterns
As I look back on ISA Automation Week 2012, there is a lot A long term collaborator of ISA-FPID and very enthusiastic
to take away rich technical sessions, interactive tours at volunteer at ISA has recently published a fascinating
one of the most dynamic properties on the planet, article in Control Magazine. You might have recognized:
productive and thought-provoking standards meetings Dennis Brandl.
but one thing that really stands out to me is all of the
connections people made during that week. You might remember his 2006 very popular book: Design
Pattern. In this article Dennis reminds us that: Good
One of the main reasons that I joined ISA back in 1993 was software design solves hard problems in a standard way
to make connections with people in my field. I wanted to and is well documented and easy to maintain and extend
have intelligent conversations with people around the which was part of the message in the book.
world who shared my challenges. I wanted to ask
questions, share my knowledge, and figure out better ways However this article focused on as opposed to code that
to do my job. I have found all that and more at ISA, and seemed reasonable at some time but has spread and
every time I see a group of people getting that same benefit become a nightmare to maintain or extend which he calls
from our Society, I take notice. See, I think that making Anti-pattern.
connections between people with similar interests and
experiences is one of the biggest benefits of being part of Anti-patterns are often the root cause of the 20% of a
ISA, and an event like ISA Automation Week is the perfect system that cause 80% of the problems. They come from
place to dive in. attempt for software designers to circumvent conflicting
requirements or design specification by create very
Walking through the halls a few weeks ago in Orlando, I specific code to manage the exception. As the system
overheard conversations about dozens of interesting evolve, the exception manipulation become obsolete and
topics: cyber security, compliance issues, safety in plants, very often unnecessary.
the challenges of developing people to meet the needs of
our plants long-term, the importance of alarm He proposes very interesting solutions in a very
management and the list goes on and on. I can only reasonable context, but I dont want to spoil the end so I
imagine the implications of the conversations I overheard, suggest you have a look at:
but I would bet that someone came away with a new http://www.controleng.com/channels/manufacturing-
solution to try in their plant; I would bet that someone it/manufacturing-it-news/single-article/design-patterns-
went home feeling like they really helped a colleague; and I and-anti-
would bet that many people are continuing those patterns/571eeda95dd049436294a12b220feb38.html
conversations right now. That is a primary function of this
Society: to bring people together, facilitate conversation, For more information on Dennis and his work , he can be
and, as a result, bring out the best in each of us. reach at the address in the officer list at the end of this
newsletter.
One attendee, Eloise Rouche of Dow Chemical Company,
gave this quote for our press release that was sent after
the event: The networking opportunities were fantastic.
Sharing information and participating in in-depth
conversations about critical topics like process automation
safety made the late nights and bleary-eyed mornings
worth it.

This is a great way to explain one of the biggest take-home


benefits of ISA Automation Week: there are a lot of late
nights and hanging out with old and new friends, but
conversations that start there are priceless. Thanks to all
of the attendees for making my 2012 conference
experience one that Ill remember and rely on when I
get back to my day job.

P.S. Do you have comments or feedback about ISA


Automation Week 2012? If you were there, and want to
www.isa.org/standards or contact Charley Robinson of
ISA staff, crobinson@isa.org .
Alarm management technical
report published
Alarm Management for
ISA-TR18.2.6-2012, Alarm Systems for Batch and Discrete Food/Pharmaceutical Processes
Processes, is published and available at
By Joseph Alford
www.isa.org/standards. This technical report covers the
application of alarm management principles in ISA-18.2-
2009, Management of Alarm Systems for the Process Introduction
Industries, to batch and discrete processes. An ISA webinar was presented by Dr. Alford on Oct. 11 on
Alarm Management for Food/Pharmaceutical Processes.
The alarm system serves to alert, inform, and guide Slides used in the webinar are available upon request. An
operators regarding abnormal process conditions or invitation was subsequently made by Division leadership
equipment malfunctions. It may include the basic process to create an article based on some of the more advanced
control system (BPCS) and the safety instrumented system aspects of alarm management that were presented in the
(SIS), each of which uses measurements of process webinar.
conditions and logic to generate alarms.
cGMP Expectations Regarding Alarm
The general principles and techniques described in the
technical report are intended for use in the lifecycle
Management
management of an alarm system based on programmable Since quality cannot be tested into product (e.g., some
electronic controller and computer-based human machine product quality tests destroy the product sample), cGMP
interface (HMI) technology. Use of this technical report expectations are for manufacturers to operate a validated
should consider batch and discrete process alarms from all process in which consistent high quality product is
systems presented to the operator, which may include produced. This implies that those process variables
basic process control systems, annunciator panels, safety potentially affecting product quality (known as critical
instrumented systems, fire and gas systems, and process parameters) be monitored, with process
emergency response systems. (See the May/June 2012 deviations associated with these variables formally
InTech Standards column for more on batch and investigated as part of forward processing and marketing
continuous applications.) decision making.

Following the recommended guidance in this technical In order to know when a process deviation occurs,
report will not necessarily ensure that alarm management scientists developing a product using a particular
problems will be avoided. But it will help to identify and manufacturing process must determine the Proven
address alarm specification, design, implementation, and Acceptable Range (PAR) of each CPP. This is usually
management opportunities that are important to batch determined by running pilot plant experiments during
and discrete processes. It can also help minimize the product/process development. The philosophy is that
generation of nuisance alarms that could complicate and operating a plant such that all CPPs are maintained within
frustrate an operators awareness, understanding, and their PARs maximizes the probability that acceptable
response to abnormal situations. quality product will be produced.

ISA18 has five other working groups producing technical Regarding process alarms, there is a natural tendency for
reports to augment ANSI/ISA-18.2-2009, Management of automation engineers to set CPP alarm limits (i.e., alarm
Alarm Systems for the Process Industries, commonly setpoints) at the PAR values. The problem with this is that
referred to as ISA-18.2. These reports will add rationale, the components involved in making process
usage guidelines, and examples in different areas of alarm measurements (sensors, transducers, etc.) contribute to
management. The other five reports are: some finite uncertainty around the measurement. E.g.,
some flow measurements have a vendor stated
WG1 Alarm Philosophy uncertainty of 1 % of span. A consequence of this is that a
WG2 Alarm Identification and Rationalization measured value that equals a PAR value may or may not
WG3 Basic Alarm Design represent an actual deviation, depending on the
WG4 Enhanced and Advanced Alarm Methods measurement uncertainty involved (MU). Therefore, an
WG5 Alarm Monitoring, Assessment, and Audit important consideration for automation engineers is to
consider MU in setting alarm setpoints- as sometimes the
The committee hopes to follow this technical report soon result will be to set alarm setpoints inside the PAR values.
with those from WG4 and WG5. For more information on These concepts are illustrated in Figure 1.
this and other ISA18 technical reports, visit
identical (variability is usually significant for
bioprocesses), acceptable values for cell mass at a
particular point in time fall in a specified range rather than
being one precise value.

So, the idea in developing Fig. 2 is to identify several (e.g.,


20) acceptable successful historical runs of a particular
fermentation process, then calculate the time varying
average value of a targeted PV (e.g., cell mass) and also
calculate the +/- two standard deviations around the
average value. The +/- two std. deviation plots are then
Figure 1, Setting alarm limits (setpoints) as a function of PAR values copied (or piecewise fit to polynomial equations) and used
as a background overlay on the operator console trend
Special characteristics of batch processes plot. What the operater then sees is the current PV plotted
Batch processes differ from continuous processes in against the background range of known successful batch
several respects, some of which are important runs. It is then easy for the operator to see that the run is
considerations in specifying alarm system functional proceeding normally, without having to memorize or refer
requirements. E.g., most continuous processes have one to charts as to what the expected PV acceptable range is at
long term operational phase, and are linear, steady-state a particual time. If the automation system is programmed
processes. Alarm attributes usually are not changed during to allow alarm setpoints to follow the value of the +/- two
the operational phase of the process and alarm std. deviation plots, then an alarm can be generated when
displays/records are presented using calendar time. the PV exceeds this range. In the case of Figure 2, an alarm
Central control rooms are normally manned 24 x 7 so a is generated at abut 13 hours into the fermentation when
control room indication of an abnormal situation will the cell mass is higher than that associated with historical
quickly capture the attention of an operator. Batch successful runs.
processes usually consist of several shorter operational
steps/phases, most of which are time-varying, non steady- The abnormal situation shown in Fig. 2 is relatively
state and non-linear. Users desire that alarm common in industry, occuring when a fermentation
displays/records are presented in relative time (time since becomes contaminated (e.g, with foreign bacteria). In fact,
the beginning of the batch lot or lot step) and need for data Fig. 2 represents one of the most useful techniques
and alarm records to include important batch identifiers available for early on-line automated detection of
(e.g, lot number, batch step, alarm class) to assist in alarm contamination.
record sorts, queries, and batch record generation. Batch
processes are normally only semi-automated, with many Note that Fig. 2 is specific to the fermentation step of the
manual field operations still required- so central control overall process, with the X axis showing realtive time
rooms are not always manned. Means of alerting operators since fermentation inoculation (i.e., the beginning of the
outside the control room of any new abnormal situations is step). This avoids operators and process engineers having
often required. to deal with more traditional calendar time plots for which
time and effort are needed to associate a calendar time to a
It is important for project teams to consider some of these particular process step and time within the step. Note also
nuances of batch processes when specifying functional the need for near continuous changes to the alarm setpoint
requirements (near the beginning of automation projects), (often needed for some batch PVs). Plots such as Fig. 2,
since many commercial automation systems do not while highly effective and driven by the needs of operators,
provide, in their off the shelf products, all the alarm are not typically available in most commercial automation
functionality that users may desire. Use of third party systems. Figure 2, e.g., was made available in
products or customization of the automation system manufacturing plants via use of a 3rd party product and
incorporating advanced/enhanced alarming techniques some customization. So, this is a good example of the need
may be needed. to identify user alarm requirements near the beginning of
a project, when time is available and resources can be
An example of one effective batch process alarming committed to pursue any special functionality needed.
solution, implemented by a major pharmaceutical
company, is shown in Figure 2. The process is a
fermentation and the PV is a measurement of time varying
cell mass in the bioreactor (usually available via a direct
measurement or by a virtual sensor utilizing calculating
oxygen uptake or carbon dioxide evolution. The value of
cell mass during the first part of a fermentation usually
follows an exponential growth function, but then changes
due to a variety of factors in the fermentation that affect
growth rate. Since not all fermentation batches are
There are several reasons for lack of significant alarm
record mining in data historians. One is the lack of good
English descriptors in alarm records that are
understandable to the technical service (TS) and quality
control (QC) scientists that often have need to review
alarm information. E.g., describing an alarm as associated
with FIC101 may make perfect sense to the automation
engineer configuring the control loop and associated
alarm, but is Greek to the microbiologist scientists who
review CPP data and alarm records in order to pursue
required deviation investigations.

Another reason is the lack of sufficient identifiers in alarm


Figure 2, Using Historical Run Performance to Determine records to facilitate quick and efficient sorts and queries.
Alarm Limits for a Fermentation Batch Process E.g., in preparing a batch report (either manually or
automatically), TS and QC scientists will often want only
those records associated with CPP alarms for a particular
Mining Batch Data and Alarm Records batch lot. If batch records dont have this information in
The good news is that most food and pharmaceutical the record tags (or automatically available via other
companies have now collected a large amount of data means), then such activities are tedious, boring, and an
regarding their manufacturing processes due to advances inefficient time use of valuable resources.
in recent decades in computer systems, sensors, data
storage, and cost reduction. E.g., process control systems
deal with updates to many process variables every second, Helping TS/QC with deviation investigations
use them to calculate additional variables, with much of An example of a graphic developed by a major
this information potentially sent to the process historian, pharmaceutical company to help TS and QC personnel in
either in raw, filtered, or compressed form. Additional investigating deviations is that shown in Fig. 4. As noted
information comes in at lower frequency from at-line and before, all CPP deviations must be formally investigated for
lab assay equipment. processes subject to cGMPS. Further, all deviations should
be investigated anyway as a best practice to discuss ways
The bad news is that most of this data continues to sit in of keeping the deviation from happening again.
historians as just data and is not mined for the nuggets of
gold representing its information and knowledge content. Rather than having someone spend time gather the
Alarm records are a good example of this observation. pertinent PV trend plot, alarm records, and any operator
Plant automation systems generate large numbers of comments associated with a deviation, and then trying to
alarms and associated alarm records (often thousands per correlate it all, time wise, on a graphic for presentation to
day) but little use is typically made of these records. The the deviation investigation team, why not have the
situation and opportunity are captured in Figure 3, which historian automatically generate the desired graphic. In
shows data (including alarm records) residing at the lower the Fig. 4 example, one plot shows the pH trend plot which
level of the triangle. The upper levels of the triangle show includes the deviation incident, and then shows on the
some of the opportunities to mine the information and same time scale when the alarm activated, when the
knowledge content of these records. operator acknowledged the alarm, and when the alarm
deactivated. It also lists additional detail of these discrete
events along with the comment that the operator entered
into the system explaining what he/she did in dealing with
the abnormal situation. A key question that this plot helps
management with is in determining whether operator
actions were timely and appropriate in dealing with this
deviation.

Figure 3: Moving from data to information and knowledge


Batch Reports, Lot History, Master Recipes, etc. Therefore,
for plants pursuing an integrated computer environment,
there can be a need to sort and send selected data (in
lower level systems such as the historian, alarm loggers,
and 3rd party products such as expert systems) up to MES
systems to provide input to lot history data/knowledge
bases and batch reports. Communication may be needed in
the opposite direction as well, as process control systems
may need to access higher level systems to obtain Master
Recipes (containing configured alarms) as well as LIMS
systems to obtain the latest values of lab assays for
ongoing manufacturing processes.

A second significant challenge exists due to the


proliferation of manufacturing unit operation equipment
from different vendors, with each mounted on a skid and
Fig. 4, Combining deviation information from various sources controlled with an equipment vendor supplied embedded
system (usually a PLC) utilizing proprietary control, HMI,
Challenges with alarm management and alarm logic (ISA is calling these systems Packaged
While many aspects of alarm management are well Systems). These equipment vendors are usually reluctant
understood and documented in such references as the to customize their code to meet specific customer
ANSI/ISA 18.2 Standard, there are still challenges in requirements as they then need to support multiple
creating a unified effective alarm management versions of their software. Further, their systems are often
environment for operators. Figure 5 is shown to help not compliant with the ANSI/ISA 18.2 Standard on Alarm
describe two of these challenges. Management and are usually limited in the data and
information they are capable of exporting to external
computers (such as a central plant DCS). This creates a
situation where operators need to learn about and deal
with several different alarm systems in a plant, with the
rules of navigating through alarm displays, alarm
acknowledgment, etc. different for different equipment.
This is a very undesirable situation from an operators
perspective and is also counter to alarm philosophy
objectives as stated in the ANSI/ISA 18.2 Standard.

ISA is currently in the process of creating a new 18.2


working group (WG7) to deal with the packaged alarm
system challenge.

Conclusion
This article touches on a few aspects of alarm management
of special interest to practitioners in the regulated food
and pharmaceutical industries, dominated by batch
processes. It is hoped that these aspects of alarm
Figure 5: Distribution of plant automation systems involved with management convinces readers of the need to think
alarm management through needed alarm functionality, as well as to how
alarm records will be utilized, near the beginning of an
The first challenge recognizes that personnel in modern automation project. In so doing, adequate time can be
plants are working to integrate enterprise and process available to consider 3rd party products or any needed
control computer systems, supported in part by the ISA 95 customization of the alarm system. Users should NOT
standard which defines 5 different levels of automation. assume that off-the-shelf commercial systems will
As computer experts are aware, it can be difficult to provide all the alarm functionality they may need in their
communicate and share data and information between plant.
systems, as such information often exists in different time
frames, formats, and contexts in the different systems.
Referring to Figure 5, some alarm information may exist at
Level 3 in what is known as the MOM environment
(Manufacturing Operations Management) which in turn
consists of Manufacturing Execution Systems, Laboratory
Information Management Systems, Maintenance Records,
FPID at Automation Week 2012
As usual, most divisions have held a division meeting
during automation week. FPID was no exception. Led by
our Director, Alex Habib, several of our division officers
reviewed last year performance and planned for the year
to come.

During the last year we have gained several new members


and we are now over 1,500. We have provided a FPID
speaker from Genentech to the analytical division
symposium at Anaheim, CA.

We have started establishing a strong relationship with


ISPE (International Society of Pharmaceutical Engineers)
chapters in:
North Carolina
Ireland
Philadelphia
New Jersey

We have issued several newsletters in 2012:


Winter
Summer
You are reading Fall now

We have delivered Live webinars on:


Vision Systems
Batch Standards (ISA-88)
MES Standard (ISA-95)
Pharmaceutical Automation Project Management
Alarm Management on October 11, 2012

We will start planning for a Symposium in 2014; a first


kick off phone conference was performed on October 16
to:
Develop Theme
Assign committee chairpersons
Establish location
Establish a date

If you like to volunteer please contact:


Alex Habib
FPID Division Director
Alex-habib@msn.com
Models and Support for Alarm 1. Alarms reflecting inherent safety or economic limits on
Operation; Batch vs. Continuous the plant, to be avoided, by their very nature.
2. Alarms, for which well-defined corrective operating
By E. H. Bristol ; The Foxboro Co.
procedures exist (These alarms may well fall in the first
class as well.).
ABSTRACT 3. Alarms, whose importance and meaning is situation
dependent (To the extent to which the situations are
Alarms are traditionally handled as standard contacts or detectable and recognized, these alarms may also fall in
parameters associated with process sensors, with uniform the first two classes.). Batch by its nature extends this
implementation, but no application related Alarm class.
Management organization or strategy. For continuous 4. Alarms, which generally reflect qualitative constraints,
processes, where there is a possibility of independent not well formalized. Ideally these alarm limits would be
alarmed events, in different parts of the production, this periodically tightened, as operating experience justified it,
has been an easy, lazy approach to Alarms, depending on in an SQC kind of approach.
operator experience and discretion. In this context, Alarms
are the main connection from the automation to the The first two kinds of alarm are the result of good
operator, for addressing abnormal process operation. engineering, and subject to it. They should be included
Alarms of this sort address problems that we have been with the orderly automation design, and are improved by
unable to automate. But they are often as much a source of better automation1 and better rationalization of the
consternation as help. The Alarm problem is thus twofold: operators role. The last two kinds of alarm constitute the
How to make traditional Alarms more useful, and how to greatest problem to orderly, continuous operation. And
provide a more engineered approach to Alarming. For the yet, properly supported, they may offer the greatest
former, the paper suggests simple Alarm tools requiring opportunity to profit from any operating discretionary
only the most basic configuration. These include Trending, capability. The first part of the paper will address Tools
automatically generated situation descriptions, causal designed to better support the operator in coping with
Alarm analyses, and situation adaptive Alarm operation. large amounts of this kind of Alarm information, most
But Batch control particularly lends itself to better useful in predicting and diagnosing abnormal process
engineering of the relative role of the operator and states.
automation because the production steps follow a natural
progression with expected normal and abnormal Excepting a few alarms with well-defined (but not
consequences. In this case, Alarms and their response can automated) operating responses, most of the alarm
be integrated into the natural sequencing. activity will thus be coping with deficiencies in the process
automation and operator understanding, and in the
adjustment of the alarms themselves. The Alarm Analysis
INTRODUCTION Tools must minimize these deficiencies, but they will
Process control alarms are the main automation system function better with better application automation. They
vehicle for dealing with operations too abnormal for presume a general Process Control vantage point, neither
realistic automated accommodation. These abnormalities Batch nor Continuous.
will often be unfamiliar to the operations and applications
personnel. Some will be very mild, some catastrophic. The Tools are optimized about poorly understood process
Large continuous applications may have thousands of characteristics, requiring easy sensitivity to unusual
potential alarms defined and hundreds active at any point behavior, and rapid exploration and diagnosis.
in time, sometimes within one or more causal showers. Accordingly, they should not be overloaded with
Traditional alarms, however well designed, are treated as configuration or operating features. In particular, the Tools
limited sensor attributes, rather than coherent control must avoid complex, configured, alarm specific responses
elements. They are inherently inadequate to support the more appropriate to well-understood, systematic plant
operator in recognizing and coping with significant automation. The related configuration of the alarms and
abnormal or crisis situations. The push to operate with their associated displays should be governed by
fewer operators places that much more of a load on the generalized display Policies applicable to groups of alarms
existing ones, exacerbating the problem. The initial tool rather than by detailed configuration of the individual
discussion focuses on improving the analysis and alarm action. The display approach must support quick,
presentation of this traditional alarm information. The easy, transparent operator action.
later discussion of Batch control discusses how this kind of
operation naturally encourages a more rationalized Processes with more systematic relations between process
approach to Alarms. events and possible alarms support easier formalization of
alarms and the operating role, perhaps with traditional
Alarms fall in a number of classes, depending on the extent automation, perhaps with Expert Systems. For example,
to which their role is well understood: the large continuous plant has all aspects of production
running at the same time. The alarms are based on Styrene Plant, Feed Tankage, Furnace, Heat Exchanger,
individual process variables. Event/Alarm relationships Reactor, and Separator buttons with their graphics).
may be localized to the part of the process in which the
causal event may occur. But multiple trains of events may Categorizations, which are top level nesting Categories
occur at the same time in different parts of the process, whose subCategories collectively contain all of the
merging together in an overall alarm flood or shower. On application alarms. They represent different ways of
the other hand, as discussed in the later part of the paper, completely dividing the process alarm variables. Knowing
the events of a batch plant are ordered, sequentially in the Category locations of an alarm in each of the different
time, giving rise to greater predictability in any associated Categorizations is like knowing the coordinates of a point
alarms. This simplifies operator diagnosis and the in space. (The # marked buttons, in the Access Map,
development of standard procedures. It also permits correspond to unnamed Categories whose alarms fill out
alarming of specific events independent of generalized the associated Categorization.) The Categorizations
alarming standards. support the hierarchical display access. They also support
the Mask/Filter, active in most of the displays (e.g., the
Batch also emphasizes certain issues not normally Styrene Plant and Furnace graphics above), as will be
recognized in continuous control, in particular the defined shortly.
dynamic variation of operational conditions with changing
operational phases. The Alarm Tool Category concept is Patterns, which are Categories, whose alarms make up a
the most basic vehicle for making the operation phase symptom pattern, and whose order of occurrence, in a pre-
dependent. Most of the accommodation of the Tools to the defined time window, is presumed significant. They are
Batch environment will be through the APIs which allow divided into Cause Alarms, and Main (important) and
the construction of Batch system commands or links to Secondary Effect Alarms. In the Access Map (and Mask
control the status of associated Categories from the Batch displays), their buttons are distinguished in Italics.
procedure. Batch will motivate different attitudes for
combining the configuration of controls and Categories,
and for archiving prior Category configurations. BASIC ALARM RECORDING: LOGGING AND
TRENDING; SIDE LOGS
THE ALARM TOOLS
Traditional Alarm display takes three forms: Alarm Panels,
Indicators in Graphic Panels or Displays, and Logs. Of these
The accompanying demonstration develops a continuous
the simplest (below left) is the log, shown in its role as an
Styrene Plant example, showing an application based
absolute archival record of events for later analysis. But
hierarchical organization. [1] The demo expresses a basic
even here, and particularly as a vehicle for online
set of graphic displays including only alarm related data,
diagnosis, the log has a lot of defects:
which in real life would contain all kinds of information.
The demo also shows a number of default displays, which
The events of a given variable are spread among many log
would exist even when a graphic was not configured.
entries,
The design concept is based on simple hierarchical
The log is hard to search for different alarmed variables
Categories of alarms, extended with as many different uses
or events,
of these Categories as possible. The Access Map includes a
Process Divisions based hierarchy. But it also shows other
Events for a given variable are hard to correlate.
hierarchies as well, based on the process Streams or
Situations (in Batch: Phases) or Operations. The resulting
For its basic archival and post situation analysis purpose,
multi-hierarchy is intended to be fully user configured:
the log may be adequate. But it is hardly an operational
The alarmed process variables are user-defined, as well as
tool. For operations, the Tools provide a trending
the choices of higher level Categories in the complete
capability, which supports each alarmed variable with its
structure. Each of the demo Categories reflects possible
own (in this case letter coded) trend line. The demo codes
(but not required) user naming and content.
the six kinds of alarm events (Deviation, High Deviation,
Lo, Hi, Lo Lo, and Hi Hi) with appropriate lower or upper
case letters. More powerfully, these could be color coded,
CATEGORIES, CATEGORIZATIONS, AND giving rise to our name for this kind of trending: Battle
PATTERNS Ribbon Trending. The color-coding can also trend the
acknowledgment history. The display is intended to allow
The whole system is based on three mechanisms for user choice of time window and prediction display as well.
hierarchically classifying Alarm conditions: The log and trend figures show that the Mask/Filter
extends to these displays.
User defined Categories, which are lists of alarms (listed
in the default display for any Access Map button except the
One benefit of a trending (which groups all data about a Full DeSelection. This allows the alarms of a deselected
single variable in a single line) is that its lines can be Category to be fully masked, without regard to other
ordered alphabetically (and include only lines for alarmed selections.
variables) for easy search. This automatic ordering can be
extended to fit hierarchically grouped tags. The system Partial Selection, applicable only to Patterns, as described
automatically orders every listing, button array, or other later.
appropriate display, in every figure, consistently in this
way. These modes permit the, always visible, manual or
automatic, expression of suppressed or displayed alarms.
The demo is based on various standard full-screen The two sets of five small buttons, next to the
Operator displays covering particular classes of Select/DeSelect button control the choice of available
information calling for sustained attention. In addition it is modes. The upper set controls the modes selectable for
handy to have a few simple pull down displays supporting individual Categories. The lower set controls the modes
a quick summary capability. The figure shows what we selectable for the Select/DeSelect and Categorization
have called Side Logs, displaying different overall alarm buttons. The lower set does not allow the change of mode
views. to or from any mode not permitted in the lower buttons.
From top to bottom three kinds of display are shown: The result is that special conditions, for example an out of
service Category, can be imposed without affecting the
A Pareto ordered listing of the Alarms, indicating the normal operator use of the Mask/Filter.
frequency of every alarmed variable. Pareto ordering
supports a nice alarm discipline encouraging the
development of general procedures for addressing or ONE WORD AND PRIORITY SUMMARIES
removing frequently occurring alarm conditions.
One of the dilemmas of operator display is the meaningful
An alphabetical listing of all current alarm conditions. summary of information in alarm indicators representing
This also includes One Word and Priority Summaries groups of alarms. The figure below shows a kind of display
(defined in a later section) for the whole plant. that might normally be used in an alarm panel, displaying a
large number of alarms as labeled alarm lights. In this case,
A trending of the most recently changed alarm conditions. each indicator group summarizes the state of a particular
process unit (Category). Each indicator includes button
selection of the associated display graphic and One Word
UNIFORM MASK/FILTER and Priority Summary text displays.

The Mask/Filter is displayed as a row or column of buttons A One Word Summary is the name of that other Category
with their associated Categorization buttons. It allows whose computed intersection with the group indicator
controlled alarm display with selection of any Category in Category contains the fewest listed alarms while
any Categorization. The alarms displayed in the current containing all of the active alarm conditions of the original
(list, graphic, log, or trend) display are those which occur Category. It is thus a computed, most-descriptive summary
in at least one of the selected/highlighted Categories of of the Alarm State of the original Category. Where there is
each Categorization. Thus the selected alarms occur within only one alarmed variable with active Alarm State in the
the intersection of the different Categorization selections, Category the name of that variable with its state becomes
suppressing all others. the One Word Summary. In this case, a low level alarm
condition can, when nothing else is happening, percolate
The selection process is designed to make Category up to any higher level display. As more activity takes place,
selection or deselection easy, whether carried out the level of abstraction provided by the One Word
incrementally or from scratch. The Selected/DeSelected Summary goes up, providing the operator with a constant
button allows all Categories to be selected or deselected, information rate but the best possible description at that
defining the Selection/DeSelection starting point. Similarly level.
the Categorization buttons allow the Categorization
Categories to be all selected or deselected. Collectively the One awkward tendency in alarm processing is to configure
group and individual buttons simplify the Mask/Filter alarm priorities in neutral categories (e.g. categories 1
selection. through 5). This has the side effect of causing most alarms
to be assigned the highest category. A better strategy is to
So far the discussion has addressed Category Selection and define meaningful, user named Categories of grouped
DeSelection. The Category selection process can be alarms, as described above, and then prioritize the
extended in three more ways: Categories. The Priority Summary is the name of that
highest priority Category still containing any of the alarms
Full Selection. This allows the alarms of a selected of the original (e.g. Furnace) Category.
Category to be fully displayed, without regard to other
selections.
The One Word Summary Category will always fall in a of the Metric, are displayed with asterisk. In this form, the
single Categorization. A more powerful use of the display permits operator diagnosis and action on some
underlying concept would make use of intersections prescribed recovery procedure.
between Categories in each Categorization, to further
group and localize the alarms of the overall indicator The Pattern can be used for event prediction as well. In
Category. This is not practical in the above displays. this case the Metric is computed, based only on Cause
However, a special alarm display is provided to allow this events. When the number of such Causes exceeds a
form of analysis. specified threshold, the remaining Main and Secondary
Effect Alarms may then be predicted. The figure below
The figure shows three main display elements: shows the default listed display for the earlier Fire Hazard
(as with any other Category). In this case all occurring
An alphabetical listing of all alarm variables with their alarms are displayed with their alarm state. In addition,
states, similar to that shown later in the default displays. the remaining alarms are indicated as predicted by the
deeper shading. (The right-hand side checkbox display is
A Categorization grouped Category list (similar to the designed to show the complete Alarm State of any selected
Mask/Filter) that allows the Category membership display alarm variable, with other related information.)
of selected alarms.
The same shaded prediction is applied to any display
The Mask/Filter. including the predicted alarms. The Pattern analysis
provides a much simpler substitute for Expert System
Using the Mask/Filter and Set Buttons, any arrangement of application. But it can also be interfaced with an external
the alarm variables and states can be selected (then Expert System, through APIs, to integrate their analyses
highlighted in darker shading). In particular the alarms in consistently into the Pattern display. This supports the
any Category can be selected. The Categorization/Category best of both worlds.
list then indicates (by different shadings), which other
Categories (in every Categorization) include all of the
selected alarms, and which include only some of them. In ADAPTIVE ALARMING
this way, the display expands the One Word Summary as
an intersection of Categorizations. The display can also be Simpler that normal control configuration, these Patterns
used for a number of other What If explorations. and Categories can still become too complex for
convenient user configuration. There is a need for
automatic recognition of usable Alarm situations,
PATTERNS Categories, and Patterns. Moreover, the alarm limits
themselves can call for situation dependence, making the
A number of papers, over the years, have proposed process of setting limits still more complex than it already
showing alarms in terms of recognized sequences of alarm is. The Alarm Tools include a number of mechanisms for
events reflecting causality between the different simplifying all of this:
events.[4,5] Recognition of such a sequence can be used
for diagnosis or prediction. The simplest way is to store Those basic alarms, which do not have inherent limits (as
the sequence and have the system record occurrences of in category 1, above), will be supported by an adapting
events, which follow that order. The earlier reference system, which keeps track of worst case excursions. The
recognized that the actual occurring alarms might never engineer can use these to support semi-automatic setting
follow the sequence directly; moreover there might be of the limits.
several sequences occurring concurrently. Accordingly, a
Metric was computed which identified the longest, Categories and Patterns can be supplied with their own
properly ordered, subsequence occurring within a defined set of alarm limits so that whenever the Category or
time window. The most likely sequences could then be Pattern is selected, the special limits replace the standard
provided to the operator for his more detailed selection ones. In the case of Patterns, the replacement would occur
and analysis. in a timed and sequenced order natural to the Pattern. This
allows the Alarm Tools to support situation dependent
The left-hand figure above shows the resulting ordering alarm limits. The Categories and Patterns used this way
with the displayed Pattern Name buttons, number of would be supported by historical records, which allowed
alarms in the best match sequence, and the normalized the after-the-fact engineer supported adaptation of these
(against the total number of Pattern alarms) Metric. The limits as above.
buttons permit selecting a display for a particular Pattern
(the right-hand figure), with sequence ordered Alarm On recognition of a problematic situation the operator
variables divided into Cause, and Main (important) and could call for a special historian recording action that
Secondary Effect Alarm variables. Each variable is would record potential Pattern events for later
displayed with its current alarm state. Those alarms, which configuration as a Pattern.
occurred in the sequence order, as part of the computation
We have defined a special event based auto-correlation can build standard practices, similar to those developing in
strategy, which allowed an engineer to process months of the continuous control world.
data looking for correlated events constituting potential
Patterns. CONCLUSIONS
Alarms are the main connection between the operator and
POLICIES the automation when addressing operation of the process
outside of the normal support of the automation. Such
Currently many vendors support the configuration of operation, beyond traditional engineered automation,
alarm characteristics on a variable by variable basis. This requires tools that allow the operator to effectively call on
is particularly cumbersome for annunciation, his working experience. The paper first presents a set of
acknowledgment, and similar issues, which depend on tools, based on the simple configuration of multiple
grouped action for effectiveness. A much better strategy intersecting hierarchies of Categories. These tools allow
would be to configure such issues for different Categories, the operator to control his displays by display selection
as grouped Policies, in the same way that priorities are and information masking. The One-Word Summaries
defined above for the same Categories. The earlier Priority provide summary capability. This permits low level
discussion already gives an example of Policy information to percolate up through higher level displays
configuration discipline. in a controlled level of abstraction with high detail when
little is happening and abstraction when more is going on.
The Pattern concept supports causal modeling for
BATCH CONTROL ALARMING vs. prediction and diagnosis. Logging and trending access time
data. This support exists on top of more organized alarms.
CONTINUOUS CONTROL ALARMING Batch control particularly lends itself to these more
organized alarms, especially ones indicating exceptional
The above Tools address alarms in general terms, as events naturally related to particular phases of the
related to individual process variables, thorough the sequencing.
simple mechanisms provided by the Categories. They
particularly support continuous processes, or the
continuous support facilities in Batch processes, by
allowing the diagnosis between the many different
production activities and failure causality sequences
potentially coexisting in time. In the Batch process, each
unit or train is naturally operating in a more restricted
focus, perhaps in a single phase.
Even in the multi-train facility, the alarms of each train
correlate with the particular phases running in the train.

In Batch control it is more natural to build an alarm


structure clearly integrated with the logic already
operating in the control program, and with any normal
operating display. These alarms can be seen as special
kinds of requests extending the normal operator requests
for action. At the same time, the natural program logic
lends itself to a higher level of automatic failure reaction,
reducing the extent and dependence on operator alarm
reaction.

All kinds of processing should be better organizing their


collective manual/automated operational support to
distinguish different modes or regimes of normal and
abnormal operation. These regimes need to be related to
clearly distinct kinds of operator display and automation
support. The design of Batch control systems should take
advantage of their natural characteristics. For example, as
relates to the Tools, Batch control already has the Pattern
built into the sequencing of the control program. Thus the
first part of causal prediction and diagnosis is already
inherent in the Batch sequencing. We still need to develop
generally recognized regimes of operation, on which we
Free ISA Web Seminars Is a Fire & Gas System a Safety Instrumented
Isnt it time you explored ISAs free web seminars? As an System?
ISA Member, you can enjoy unlimited, free access to a Manufacturing System Security Using ISA SP99
library of over 40 online pre-recorded web seminars Motivating Team Members
covering todays hottest topics. Its a great way to get extra Network Security Series Part 1: Cyber Security
training that fits your schedule. Risk Assessment for Automation Systems
Network Security Series Part 2: Firewalls and
Below is an alphabetical listing of the seminars. They can Filtering Security on the Plant Floor
be accesses at: Network Security Series Part 3: Encryption and
http://www.isa.org/template.cfm?Section=Web_Seminars6&tem Virtual Private Networks
plate=/customsource/isa/ArchivedSeminars/Series.cfm
Next Generation Fieldbuses - Picking the Right
One
Dont forget, you must be logged into www.isa.org
Noise Reduction in Signal Replacement
Applying Manufacturing Execution Systems
Overview of Enterprise Control Integration
Benefits of ISA Certification
Overview of Foundation Fieldbus Technology
Boiler Control Fundamentals
Overview of HMI Applications
CSE PE Exam Review Part 1: Standards &
References Safety Instrumented Systems Series Part 1:
Determining Safety Integrity Levels
CSE PE Exam Review Part 2: Measurement I
Safety Instrumented Systems Series Part 2: The
CSE PE Exam Review Part 3: Measurement II
Pros and Cons of Logic System Technologies
CSE PE Exam Review Part 4: System Analysis
Safety Instrumented Systems Series Part 3: Field
CSE PE Exam Review Part 5: Final Control
Device Issues and Their Impact on System
Elements
Performance
CSE PE Exam Review Part 6: Control System
Sell More Through Effective Technical
Implementation
Presentations
DCS Alarm Assessment Overview
The ISA95 Object Models for Enterprise - Control
Deployment of ISA100.11a - The First Proposed
System Integration Part 2
Standard of the ISA100 Universal Family of
Wireless Networking of Sensors and
Wireless Standards for Industrial Automation
Instrumentation: Applications & Techniques
Designing & Applying Model Predictive Control
Strategies
Effective Project Team Building
Engineering Influence: Unlock the Secrets to
Getting Your Way
Fiber Optic Sensing
FOUNDATION Fieldbus H1 Field-Level Installation
and Commissioning
Grounding in Instrumentation Systems
How Can OPC Reduce Your Plant Control
Integration Costs?
Hybrid Systems Overview
Industrial Ethernet Series Part 1: Understanding
Ethernet Cabling & Hardware
Industrial Ethernet Series Part 2: Is Ethernet
Ready for the Plant Floor?
Industrial Ethernet Series Part 3: Designing
Industrial Ethernet Networks
Introduction to DCS Alarm Management
Introduction to ISA100.11a - The First Proposed
Standard of the ISA100 Universal Family of
Wireless Standards for Industrial Automation
Introduction To Patents & Patent Process
Introduction to Process Control
Introduction to Safety Instrumented Fire & Gas
Systems
Introduction to SCADA
Introduction to Temperature Measurement
Is a Burner Management System (BMS) a Safety
Instrumented System (SIS)?
FPID 2014 Conference Survey 7) Do you value continuing educations?
a) Yes continue at question 8.
As Alex explains in his introduction, we have started b) No continue at question 9.
working on our first Supply Chain Symposium for ISA
planned for the spring of 2014. 8) If you value continuous education, what is your main
To help us in planning this event, we are distributing a reason?
survey to get the pulse of the membership on the subject. a) It is a core value in my company.
The survey will cover subject like: b) It is an easy way to obtain professional development
units for my qualification.
1) What topics would make you take a day or two from the c) Both
office? d) None of the above

a) Supply Chain Cost Reduction / Performance 9) If you do not value continuous education, what is your
Enhancement main reason?
b) Line / Process Automation techniques and strategies a) My company is going to tough time and do not have
c) Product safety / product counterfeit prevention & any budget.
detection b) My company hired me because I have all the
d) Process Analytical Technology (PAT) education I need for my position.
e) Manufacturing IT, MES and Virtualization c) I am not a member of a professional organization that
f) Batch Control Standard ISA-88 requires it.
g) Manufacturing Operations Management (ISA-95) d) None of the above
h) Product Tracing & Tracking (Serialization)
i) Project Management 10) What is the method you preferred for continuous
j) Validation education?
k) Plant tour a) Webinar (go to question 11)
l) Food Safety b) Local Seminar / Conference (go to question 12)
m) System and ERP integration c) Focused short course

2) What must a seminar / workshop deliver for you to 11) I prefer Webinars because.
consider it successful? a) This is convenient to be able to do it from my desk, a
conference room and/or from home.
a) Technical Content, b) We have limited (or no) travel budgets for continuous
b) Networking Opportunity, education.
c) Business Acumen/Strategy, c) I have a busy schedule and can only do continuous
d) Other, please specify. education when nothing else is urgent. A recorded
webinar is the perfect tool.
3) How far are you willing to travel? d) Other, please specify.
a) Overnight drivable (100 miles)
b) Plane trip - travel capped at $1000. 12) I prefer local seminar and/or conference because.
c) Plane trip travel capped at $3000. a) If I stay at the office, I get interrupted all the time.
d) Other, please specify. b) I have difficulty to concentrate if I am on-site and
have other things on my mind.
4) If you get the information, you needed what would be a c) We have a limited budget for continuous education
reasonable fee for the conference? but if the course does not involve any major cost such
a) $50 to $100 as air travel or hotel, it is usually approved.
b) $100 to $300 d) We have a no travel policy at the site.
c) $300 to $500 e) Other, please specify.
d) $500 to $1000
13) Would you prefer to participate in a multi-week Lunch
5) If you get the information and attend one or more training & Learn Seminar series instead of going to an offsite
seminars, would be a reasonable fee for the conference? seminar?
a) $50 to $100 a) Yes
b) $100 to $300 b) No
c) $300 to $500 c) Dont know
d) $500 to $1000
14) How long should each session be?
6) Does your company encourage continuing education? a) Less than 1 hour
a) Yes b) Less than 2 hours
b) No c) Less than day
c) Sometimes d) Not important
Yvonne Duckworth CRB Consulting Engineers
15) How many weeks should the program consist of?
Kleber Falcao
a) 2 weeks or less
b) Not more than 2 to 3 weeks Elias Ferreira
c) Not important Forest Gardner
Lethicia Goncalves
16) What time of day / night (Eastern Time)?
a) In the morning Richard
b) At lunch (12:00) Hargenrader
c) In the afternoon Richard Harris
d) At night (after 5:00 PM, I understand this would Nagaraj Hegde
limit participating from Europe )
Chris Heng Cargill Inc
17) Would you pay a minimal fee for such a webinar series? Diane Heurtin HIPP Engineering and
a) Yes Consulting Inc
b) No Hugh Jenkins
c) Maybe, depend on the content. Victor Osvaldo Jesus
A separate invitation with a website link will be sent to you S. Kester
in the next couple of weeks. If you have any comments or Jeffery Kibalo
suggestions on the survey contact Alex-habib@msn.com Robert Koneck
Thomas Kosse Vipa USA
Welcome New Members of ISAs Robert Lesutis
Food & Pharmaceutical Division Lucas Lima
Scott Little Tropicana Products Inc
Name Company Luiz Augusto Longaray
Andrew Almendares Sycal Engineering Inc James Lustig
Loic Alpha Arkadiusz Maciak Wyeth Ayerst Pharmaceuticals
Boyko Baharov Signum Soft Na Elkin Manrique
James Baillargeon Medimmune Lee McDaniel
Michael Baldauff Fujifilm Diosynth Biotechnology Thomas McDonnell
Timothy Barz S. Mecum
Carlos Bedia Keld Moeller
Richard Bellelis JP O Riordan
Pat Boner Randy Oakley
Diego Caceres Eoin O'Brien Tandem Project Management
Richard Carpenter Dermot O'Callaghan Enterprise System Partners
Limited
Yamei Chen Eli Lilly and Co
Danilo Oliveira
Donald Cherry The RMH Group Inc
Jean O'Reilly Roche Molecular Systems
Xavier Climent
Lida Paz
Didier Collas Invensys
Wendell Pearson
Henry Collins
Alessandro Pestana
Chris Cook
Jaideep Raje
Gabriel Coutinho
Joseph Rakos
Brinton Crawford
Donald Richmond
Colm Cronin Leo Pharma
Paul Roberge RTI Products
Phillip Custer Instrument Technologies
Pedro Henrique Rodrigues
Jospeh DeBitetto Fougera Pharmaceuticals Inc
Bryan Rose
Adam Dittbenner Interstates Instrumentation
Ralph Russotto Corrosion Fluid Products Corp
Alexandre Djordjevic
Giovanny Sabogal
Atila Santos
David Adler
Richard Sara Special Assignment
Timothy Shannon davidadler@comcast.net
John Campbell
Katia Silva Petrobras Industry & Science Department VP
Jensy Solano Correa campbjr@msn.com
Renan Souza
Mike Nager
Bruce Summerfield Sales & Marketing Chairman
Luis Fernando Susin PUC PR Phoenix Contact Inc
(717) 944-1300
Zaida Tiruganya
Ingrid Valasco Torres Dennis Brandl
Special Assignment
Brian Vincent Genzyme BR&L Consulting
Michael Walsh Eli Lilly and Company (919) 852-5322
DnBrandl@BRLConsulting.com
Corey Williams Pharmaceutical Calibrations
and Instrumentation Llc Amit Desai
Daniel Xavier Education Chairman
Bayer HealthCare LLC
Carlos Garcia
(510) 705-7309
Christopher Hogan Parsons amit.desai.b@bayer.com
Peter Steimel Total Instrumentation &
Sunil Mehta
Programming Services Inc
Special Assignment
Zubin Varghese Ingersoll Rand PCS Engineering
Mack Hu Dalian Huaqi Electronics Tech (919) 359-4990
Co Ltd Sunil.Mehta@talecris.com
Ken Jones
Kevin Dignam
Wayne Lee Genentech Special Assignment
Michael McInroe Rockwell Automation DPS Engineering & Construction Ltd
353- 21 7305000
Michael Waker kevin.dignam@dpseng.com

Bob Hubby
Facilitator
(508) 349-1050
2012 Division Officers bob.hubby@verison.net
Alex Habib, PE
Division Director James Bouchard
Industry & Science Department VP-Elect Historian
alex-habib@msn.com jamesb@aei.ca

Mike Baldauff Rodney Jones,


Director Elect Staff Contact
michael.baldauff@merck.com (919) 990-9418
rjones@isa.org
Andr Michel
Newsletter Editor
amichel300@gmail.com

Gary Campbell
Webmaster
gcampbellisa@cox.net

Randy Dwiggins, PE
Standards & Practices
(908) 704-9041
rdwg@nnepharmaplan.com

Bill Dugery
Section-Division Liaison
dugaryw@ace-net.com

You might also like