1. Geir Simensen
Honeywell Hi-Spec Solutions Norway

2. Strategic Opportunity
The history of seeking profitability from plant performance has been heavily weighted toward process operations.
Plant and equipment availability
Process operations
Equipment and plant availability performance has been the wild card in business performance.
Information Management

3. Hi-Spec’s Unified Manufacturing Solutions

4. To Help You Proactively Manage Your Assets At The Optimum Cost
Our Purpose
To Help You Proactively Manage Your Assets At The Optimum Cost
With Internet Speed & Reach!

5. Industry Trends Asset Trends Expertise Trends Work Practice Trends
assets must run longer (>3 years between turnarounds)
impact of asset failure can be very large
Expertise Trends
downsizing and aging workforce has caused “skill gap” at plant sites
technology is evolving faster than it can be adopted
Work Practice Trends
Asset management is moving from reactive to proactive
Formal abnormal situation management is improving operations effectiveness

6. A Look At Plant Operations
A typical Production Profile for an Asset Intensive Facility for a calendar year.
95 days
79 days
62 days
47 days
23 days
30 days
16 days
Days per Year
8 days
5 days
< 60%
Daily Production
95%
100%
Production Target set by Enterprise

7. Real Life Examples
This plant had $24.2M in lost capacity due to asset availability & incidents!
24.2M
This plant had 5.8% in lost capacity!
5.8%
This plant lost $38.5M!
And this plant lost $33.5M!

8. Factors Affecting Plant Operations
Between 3-15% in Lost Capacity is attributed to asset un-availability and incidents
Plant Operating Target
Planning Constraints
Plant Availability
Operational Constraints
Plant Incidents
Production
Management
DCS/APC/ Optimization efforts
NEW EMPHASIS!!
Asset Management
Days per Year
Plant Capacity Limit
< 60%
Daily Production
95%
100%
Manufacturing
Execution
Scheduling & ERP

9. Factors Affecting Plant Operations
Get that 3-15% in Lost Capacity Back!!
Planning Constraints
Plant Availability
Operational Constraints
Plant Incidents
Days per Year
Plant Capacity Limit
< 60%
Daily Production
95%
100%

10. How you manage the health of your assets
IMPROVE 
ABNORMAL SITUATION
RESOLUTION
ORIENT
DECIDE
ACT
You become “aware” of a problem.
You determine the cause and decide what to do about it.
You take action to fix or respond to the problem.
This is the how everybody makes decisions, whether they have formal procedures for each component, or whether they perform the elements internally. It is basically the work process that every human undergoes to go from an abnormal situation to a resolution…

11. facilitates this work process
IMPROVE 
ABNORMAL SITUATION
RESOLUTION
ORIENT
DECIDE
ACT
You become “aware” of a problem.
You determine the cause and decide what to do about it.
You take action to fix or respond to the problem.
DIAGNOS
-TICS
DECISION
SUPPORT
MAINTENANCE
RESPONSE
@sset.MAX enables each element of the work process to ensure that actions are taken in a well informed, proactive mode.
@sset.MAX facilitates this process by providing tools and philosophies behind each component.
OPERATOR
RESPONSE

12. Diagnostics Scouts IntelliScout TurboScout Loop Scout Valve Scout
ORIENT
DECIDE
ACT
CMMS
IntelliScout
TurboScout
Loop Scout
Valve Scout
Transmitter Scout
and more
DIAGNOS
-TICS
DECISION
SUPPORT
OPERATOR
TOOLS
As the first step, we first must detect a potential or existing problem. This is an actual example of a heat exchanger that we have our application installed on. The green line represents where the HEX should have been operating based on our models, the blue line represents actual performance. An alert was issued at the red line, but due to the holidays, it was not a priority of operations. 3 weeks later, the HEX’s performance was finally noticed by the operators. Once the HEX was repaired, the blue lined jump back to normal. With this ability, you can plan repairs for the best time possible and save a lot of money.

13. Diagnostics IntelliScout
ORIENT
DECIDE
ACT
CMMS
Detects asset specific problems before they occur
Rotating & Reciprocating Machinery
Heat Exchangers
Pumps
Instrumentation
DIAGNOS
-TICS
DECISION
SUPPORT
OPERATOR
TOOLS
As the first step, we first must detect a potential or existing problem. This is an actual example of a heat exchanger that we have our application installed on. The green line represents where the HEX should have been operating based on our models, the blue line represents actual performance. An alert was issued at the red line, but due to the holidays, it was not a priority of operations. 3 weeks later, the HEX’s performance was finally noticed by the operators. Once the HEX was repaired, the blue lined jump back to normal. With this ability, you can plan repairs for the best time possible and save a lot of money.

14. Principle of Intelliscout
Black box statistical model
Set of correlated variables, relationships dependent on asset behaviour
Example - Motor Driven Pump
Discharge Pressure,
Specific Gravity
Intelliscout uses statistical methods to create a model of expected asset behaviour.
It knows nothing about the physics of the particular asset in question. What is required is that a set of correlated variables that bound the asset in question is available, and that the relationship between these variables depends on asset condition.
An example shown above is a motor driven pump. Motor current, inlet pressure, discharge pressure and flow are related, and their relationships define pump performance. In this example we can also write an analytical model to define these rleationships, but that may not always be possible.
Motor Amps
RPM (if variable)
Flow, Suction Pressure, Temp

15. Intelliscout
Model predicts expected asset behaviour when in good condition
Compare predicted behaviour with actual behaviour
Use historian to read actual, write predicted
Early warning of change in performance
The model reads actual data for the set of variables chosen from your Process Control System historian - see not below.
It then generates estimates of what the variable values should have been, based on the model and writes these back to the historian. A comparison is made between actual and estimated values, and if a deviation is detected, a symptom is raised in the Equipment Health Management Alert Manager software.
For failure modes with a gradual rate of degradation, e.g. wear, fouling, Intelliscout will show a deviation from expected performance in plenty of time for corrective maintenance before a failure occurs, and with minimal impact on production. Even with more rapid failures, sufficient warning may be given to process operators to avoid the production impact of a trip, as will be shown in a case study later in this presentation.
The types and direction of variation between the actual and predicted values can be mapped as individual symptoms and lead to a probable fault diagnosis in a maintenance decision support software system: Alert Manager. An example is a synchronous motor slurry pump with a valve on its output to control flow: if the valve is more open than expected, the discharge pressure is higher than expected and the inlet pressure is normal, scaling has probably accumulated at the valve restricting flow. In this example there is no flow meter.
Note: A historian is a pre-requisite for Intelliscout, note that Intelliscout will work with non Honeywell a non Honeywell control system, all that is required is that the historian that sits on top of the DCS complies with OPC (OLE for Process Control) standard or there exists a Real time data Interface (RDI) between the historian in place and Honeyweel’s PHD - please contact the author if in doubt.

16. Algorithms Steady State model Data clustering
Training space, linear, non linear
Principal Component Analysis
Technique to find the directions in which a cloud of data points is stretched most
Radial Basis Functions
Non linear functional approximation
Intelliscout models are steady state models of asset behaviour - not dynamic models. This means that start up and shut down conditions may not be valid, and the data must be “smooth” or “well behaved” - discontinuities from step changes in process conditions may cause model errors. If there are distinct operating ranges, e.g. different crude stocks in an oil refinery, or summer / winter conditions, then this may be handled by having distinct models to handle the different operating conditions. This is analagous to representing a non linear curve as a set of linear segments. The choice of model can then be handled by a reference to another variable that represents operating states.
Data is clustered to determine linear and non linear components and establish the training space (or the range of operations for which the model is valid)
The model is based on Principal Component Analysis. Principal Component Analysis (PCA) is a technique to find the directions in which a cloud of data points is stretched most. These directions represent most of the information in the data. Knowing these directions allows us to store the data in a compressed form and later reconstruct the data with a minimal amount of distortion. PCA is used in statistics to extract the main relations in data of high dimensionality. A common way to find the Principal Components of a data set is by calculating the eigenvectors of the data correlation matrix. These vectors give the directions in which the data cloud is stretched most. The projections of the data on the eigenvectors are the Principal Components. The corresponding eigenvalues give an indication of the amount of information the respective Principal Components represent. Principal Components corresponding to large eigenvalues represent much information in the data set and thus tell us much about the relations between the data points.
Reference: I.T. Jolliffe, Principal Component Analysis Springer Verlag, 1986, ISBN or ISBN

17. Intelliscout Trained on data set representing normal behaviour
Analytical model may not be possible
Too complex
Under instrumented
Intelliscout models should be trained with process data representing the normal range of operations. In this way, model behaviour is linked to process conditions. This results in fewer false alerts (or missed alerts) due to a change in the process - which is a common problem with traditional CBM techniques.
A training range (start & stop time & date) and sample interval (1 minute upwards) is defined. The data for the set of variables selected in the model is then captured over the training range from the historian and sent via the Internet to an Intelliscout server which performs the number crunching. The model is returned back to the local Intelliscout system.
Intelliscout offers the ability to model asset performance where an analytical model may not be possible due to insufficient instrumentation (e.g. no flow meter with a pump circuit), or the asset may be too complex to reasonably define its performance with a set of equations. If an analytical model is possible, we would recommend running the Intelliscout model in parallel with the analytical model - comparing them may give useful insights into asset performance.
If there is redundant information in the variables included in the asset then it may be possible to perform sensor validation. Since the model is based on the correlations between the variables, if there is a problem with one variable only, and all the others are OK, then with redundant information this cannot be attributed to a problem with the asset, and the sensor at fault can be identified and an estimate provided for its value.

18. Performance Monitoring & Optimisation
TurboScout
Performance Monitoring & Optimisation

19. Our Technological Solutions
Gas path analysis and component matching techniques to derive component fault indices
Component matching techniques to minimise fuel consumption and hence exhaust emissions
Larson - Miller creep methodology and component matching techniques to evaluate remaining life
Principal of corresponding states for real gases

20. On-line Performance Monitoring
Gas turbines fault indices
Fault indices, alarms and complimentary trends
Plant PV
Gas turbines
Compressors
Scrubbers
Coolers
etc.
Compressors fault indices
Steady-state checks
Scrubbers fault indices
Coolers fault indices
Database of fault indices

21. Off-line Model Based Analysis
Optimisation (fuel consumption)
Database of fault indices from On-line system
Predictive maintenance
Model based analysis
Production based maintenance
Optimisation (compressor wash)

22. Loop Scout
Loop Scout performs automated data collection on plant regulatory control loops
Data is transmitted via the internet to an automated analysis engine for quick response
Analysis report quickly focuses scarce engineering and maintenance resources on real problems
Improved regulatory control performance provides a strong foundation for higher level control and optimisation
As another detection tool, Loop Scout is one of our latest offerings that allows you to make sure you have a strong control foundation. With Loop Scout, basic regulatory control loops can be assessed and analyzed, before you install advanced process control. The success of multivariable control and real time optimization depend on a solid regulatory control infrastructure.
Even More Details: This resulting study can perform a detailed analysis of a particular area of the refinery in preparation for implementation of advanced control. As we said before, the success of multivariable control and real time optimization depend on a solid regulatory control infrastructure.
Using this study, manipulated variable control problems are identified and diagnosed, such that the control engineer can take appropriate action to improve the control. Post-analysis quantifies the improvement and closes the loop on learning. The plant is then ready for step testing and implementation of multivariable control.

23. Transmitter Scout Continuous monitoring and tracking of transmitters
ORIENT
DECIDE
ACT
Continuous monitoring and tracking of transmitters
Provides diagnostic information based on soft failures reported from the transmitter
Notifies when transmitter has been replaced or moved
Maintenance
DIAGNOS
-TICS
DECISION
SUPPORT
Operators

24. Valve Scout
ORIENT
DECIDE
ACT
A Predictive Maintenance Solution for All Control Valves … Even Analog.
Provides diagnostic information based on stroke travel and cycle thresholds
Reduced costs through predictive valve maintenance
Support of expanded asset management capability
Improved productivity of maintenance resources
Maintenance
DIAGNOS
-TICS
DECISION
SUPPORT
Operators

25. What Tools Do You Use for Maintenance?
Oil Analysis
Infrared Thermography
Data Historians
Excel Spreadsheets
Open Control Systems
Intelliscout
Loop Scout
Turbo Scout
MathCad
Vibration Monitoring
Process Simulation
Calendar
Traditional Condition Monitoring tools make little use of Process Control Data. However, this is one of the richest sources of information on how assets are performing. The process control system already monitors flows, temps, pressure, currents; most things that characterise asset performance. However this has remained in the realm of process engineers, while maintenance and reliability people have relied on their more traditional sources of information shown in the slide above.
Process data can be made available for condition monitoring at little incremental cost using Intelliscout. Instrumentation already exists for process control purposes.
Manual Calculations
Motor Current Analysis
Computerized Maintenance Management System
Ultrasonic Noise Detection

26. Integrated Decision Support
ORIENT
DECIDE
ACT
Maintenance
DIAGNOS
-TICS
DECISION
SUPPORT
Operators
Alert Manager
Receives Symptoms generated by the scouts + all other sources like vibration monitoring etc.
Presents user with asset-specific information to act on, complete with recommendations and automated responses.
Notifies the field maintenance staff and directs them to possible reasons, solutions, and supporting documentation.

27. Integrated Decision Support
All assets
All information for each asset
Diagnostics
Documentation
Fault history
Recommended actions
Provide access to asset specific information
Integration to existing systems
Control System
CMMS / Reliability tools

28. Structured Troubleshooting
Fault files x
Required evaluation
Evaluation done, nothing found
Symptom found to exist
Provides information for current or potential faults
Evaluation activities: Obtain current status of the symptoms key to isolating the fault.
Repair activities: Activity to repair a confirmed fault.
All symptoms defined for the asset.
Status of failures: “?” potential, “!” confirmed and “X” eliminated. Related symptoms shown on expansion.
Provide information for diagnosing current conditions.
Fault History for This Asset: Status and history of faults.
Faults for Assets of the Class Diagnosis: List of current and past faults for similar assets.

29. x Required evaluation Evaluation done, nothing found
Symptom found to exist
Alert Manager continuously monitors asset data, and when a symptom is recognised, it looks for the presence of absence of related systems to diagnose a fault. It handles the data overload problem and produces meaningful information for the maintenance practitioner: a diagnosed fault and corresponding Work Order Request to the Computer Maintenance Management System (CMMS).
This view shows the contents of the Fault Folder. The possible faults for this pump are: cavitation, Coupling fault, Gland failure, Gland leak, Impeller installed backwards, Impeller rotating in the wrong direction, Inboard bearing fault, Internal wear, Motor DE Bearing fault, Motor NDE Bearing fault, Outboard bearing fault, Out board Gland Leak. Each fault is diagnosed when Alert manager recognises the presence and/or absence of a related set of symptoms.
The fault -symptom trees for each asset are created using a simple drag and drop function. AND, OR and NOT functions are provided, so symptoms can be combined in any logical combination to produce a fault.
In the example above, Cavitation is diagnosed when Condition Set 1 OR Condition Set 2 OR Condition Set 3 is true. Condition Set 1 is true when the pump has low flow and the Inlet Suction Pressure is low AND the Pump is off its curve AND Intelliscout has given a cavitation alert AND Intelliscout has given a Discharge Pressure Alert.
Ideally, the input for the Alert Manager Fault -Symptom trees is the result of your RCM analysis or your FMECA. In this way, what you have learnt from this analysis is used in day to day diagnostic and decision support by all your maintenance & reliability staff - it does not merely sit on the shelf as a report, or reside in an offline database. In this way Alert manager operationalises the results of your RCM analysis.
Similarly, Alert manager complements your CMMS. Your CMMS schedules maintenance tasks, respources, tracks inventory & costs. What your CMMS needs as an input is what to schedule when. Alert Manager provides this information on condition by requesting work orders when it has diagnosed a fault.

30. This slide shows the Fault-Symptom tree for Internal Wear, which is diagnosed when The pump is off the curve AND Intelliscout motor current estimated values have deviated from the actual values AND when Intelliscout Discharge Pressure estimated values have deviated from the actual values.
The above views show the contents of the Fault-Symptom trees. When an actual symptom occurs it is represented with a red tick against it. When a symptom is seen, Alert Manager looks for the presence or absence of all other symptoms related by Fault-Symptom trees. If the absence of a symptom is confirmed, it has a grey cross against it. If the presence or absence symptom cannot be confirmed then a magnifying glass is shown against it, meaning please check. This will usually be accompanied by an or pager message from Alert Manager requesting that the appropriate check be carried out. This can prevent work order flooding due to an accumulation of minor tasks. Alternatively, a work order request can be produced to check for presence of a symptom.
Also shown in this slide is the folder containing the fault history for this asset. The start time shows when Alert Manager reported the problem. The end time is when the asset has been set to normal in Alert Manager. Often, records in a CMMS will show when WO request was raised, or when an activity was started or finished, but will bot related directly to times when production was affected. The on line diagnosis by Alert Manager means that the Alert Manager Fault history may be a better tool to assess production impacts of equipment failures. This is often a poorly measured but vital metric in assessing the progress of a maintenance improvement program, or the effectiveness of a maintenance department.

31. Simple Configuration Diagnostic Builder Create Asset Types
Build Symptoms
Build Faults
Build Symptom & Fault Trees
Configure the symptom & fault notifications
, Pager
User Alert
Work Order
How does all of this work? With the Asset Builder, you are able to leverage your vast amount of expertise as well as that of the asset manufacturer. The Builder lets you easily define symptoms, faults, build the relationships as well as determine the notifications.
One a particular asset type has been configured, it can be used over and over again, for example motors, pumps, compressors, etc.

32. Maintenance Management
ORIENT
DECIDE
ACT
Available via our Partners
Fully integrated with Decision Support
Work orders triggered automatically
CMMS
DIAGNOS
-TICS
DECISION
SUPPORT
OPERATOR
TOOLS

33. Benefit - Higher Availability
EXAMPLE:
Higher Availability Through:
Early Awareness,
Effective Troubleshooting,
Efficient Resolution
Fuel consumption increased
Compressor moves toward surge due to lack of flow
Compressor could not maintain flow
Gas too hot, Volume larger
Excess turbine creep life consumed
Cooler Fouled
Compressor starts to recycle gas
Flash Gas Compressor, throughput $290K day

34. Scenario… No symptoms detected
Display of controllers, valves and transmitters indicate normal

35. Orient
Valve Scout detects a variation in the travel profile of 77FV211
Alert Manager automatically executes a Loop Scout Detail on 77FC211
Recommended action : Run Loop Scout Expert.

36. Decide
Loop Scout Expert confirms valve stiction and discounts poor tuning

37. Act
Alert Manager automatically generates the Work Order for valve maintenance on 77FV211.

38. Forventede muligheter med Alert Manager for Statoil Kvitebjørn (1 av 2)
Sikre/forbedre regularitet
Tidlig feildeteksjon, “Alert” via / SMS
Operatørstøtte for rask konsis feilhåndtering
Avlasting av personell på plattformen
Kontinuerlig overvåking av mange utstyrskategorier automatisert
Standardisering av arbeidsrutiner og fokus på arbeidsprosesser
Enkel tilgang til all informasjon (prosedyrer, manualer, leverandørinformasjon etc.)
Intergrering av all tilstandsovervåking
Forenklede / ferdigutfyllte arbeidsordre / notifikasjoner
Delvis automatsk tolking av resultater
Automatisk genererte funkjsonstester og rapporter

39. Forventede muligheter med Alert Manager for Statoil Kvitebjørn (2 av 2)
Samle og systematisere erfaringsdata og feilhistorikk
Forenkler feilsøking og er grunnlag for analyse av tilltak
Forenkler deling av oppgaver mellom landbasert personell og plattform
Landpersonell utfører tilstandsoppfølging og utsetdelse av arbeidsordre
Automatisk varsling til landbasert personell fra AM
Forenkler deling av oppgaver i et multidisiplint team
Bevisstgjøring av teknisk tilstand
Stimulere til målstyring
Styring av vedlikeholdsprogram
Reduksjon i forebyggende vedlikeholdsaktiviteter