Generically, “fieldbus” describes a communication protocol. Foundation fieldbus uniquely adds the user layer. Other fieldbus protocols do not have a user layer.
In the late 70’s and 80’s, we started out with digital microprocessor-based DCS’s, which revolutionized our discipline.Not long after, “smart” devices (transmitters and valve positioners) appeared, and it didn’t take long for users to ask them to be integrated. After years of proprietary integration, ISA convened the SP-50 committee to devise a standard. It is the work of this committee that grew in WorldFIP and ISP which later merged to form FF.It was the USERS that wanted a USER LAYER. Vendors were not crazy about it.The “User layer” – device-resident standard function blocks – Are a unique feature of foundation fieldbus.AI, AO, DI, DO (MAI, MDI, etc.) are the minimum required to use data from FF in the host. Even they are configurable to do scaling, alarming, square root extraction, etc. outside the DCS. A lot of power already. In addition to the PV (process variable) they also pass along signal status every time they’re executed.ARTHM, ISEL, SGCR, INT and related FB’s are available in many devices – more chances to off-load computing tasks from the DCS using standard FB’s.PID, CSEL, SPLTR (PID control, control selector & splitter) are among the blocks used for CONTROL IN THE FIELD, a feature end users wanted to achieve truly distributed control . . . ASK – what other control system is vendor-independent and totally distributed? (answer = pneumatics)ASK – Why didn’t we like pneumatics? (answer = sloppy, unreliable, lack of precision)
Why do we tolerate “many eggs in one basket” HOST DCS of today?Ask what, if anything, single-loop integrity means to attendees“Functionally” meaning controller, operator interface, network management, historian etc in separate boxes.ASK: Can anyone comment on the number of loops per controller in their system?
"The global process industry loses $20 billion, or five percent of annual production, due to unscheduled downtime and poor quality. ARC estimates that almost 80 percent of these losses are preventable, with 40 percent largely due to operator error."ARC Insight June 10 2010Operator error? How much because of acting on invalid information, or plugged lines, failed sensors, failure to detect a “flatlining” level?
Discuss mode shedding, bumpless transfer, anti-reset windup, behavior of loops when limited, etc.We are ALREADY relying on the field devices – when your positioner is having a bad day, the “loop” will have troubles NO MATTER WHERE PID is solved.
The SAME PEOPLE that forged the robust DCS of the 80’s and 90’s were active on the SP50 committee, and infused their lessons learnt into the design of the user layer. NOT ONLY are there fewer things to fail and no dependence on the DCS or its infrastructure, EVERY TRANSMISSION (nominally, once a second or faster) communicates signal status (good, uncertain, bad, limited, etc.) and the FF FB’s do SMART THINGS by default (shed to manual, bumpless transfer) as well as providing for configurable behaviour (e.g., propagate fault forward and fault state to value).Everyone know what is meant by mode shedding? (go to a safe or stable mode, e.g. manual, on a fault)Everyone know what is meant by bumpless transfer? (smoothly transition from MAN > AUTO > CAS, for example) What is a “bump”? (it’s an upset, a process upset)Any lunatics in the crowd, do not obtain a firearm and shoot your instruments.Compare reliability of components for control in DCS (= Process Control System) vs CIF (Control in the Field)Examine the established MTTF (mean time to failure) figures for each of the components in the systemFor control in the DCS (LHS diagram) there are more partsTransmitterValveCablesTerminationsPower SupplyAI CardAO CardBackplaneControllerController Power SupplyFor CIF (Control in the Field) there are fewer partsTransmitterValveCableTerminations (<half)Fieldbus Power Supply
CIF is as reliable as the devices – which you’re relying on already.Plus, FF user layer adds in-built standard thoughtful features to further improve process integrity, availability, and robustness.Value and Status known EVERY (macro)CYCLE – e.g. once, twice, 4 times A SECOND, at the prescribed time within < 1 millisecond. NO OTHER BUS CAN CLAIM THIS.YOU JUST DON’T GET THIS WITH “POLLED” DATA!Status Propagation - GRACEFUL DEGRADATION / FAULT TOLERANCE IS BUILT-IN, STANDARD!ASK: Who would say self-inflicted screw-ups are more common than random hardware or software faults?ASK: What’s a common self-inflicted screw-up in your plant? (Bubba goes out to work on FT-88204 and instead gets on FT-88203) The SECOND he powers it down, any associated loop will shed to manual WITHIN THE SAME MACROCYCLE.Windup protection – “Limited” status propagates to slave PID, up to cascade master if implemented.Bumpless Transfer – hooks are there to prevent process upsets from mode changes.Fault state – configurable; default is “hold last position” = highest availability.
Can we apply Safety Integrity Level (SIL) calculations to evaluate basic controls? The same components make up both “loops” – why not?We asked Marszal of Kenexis Consultants to run the numbers, the same way they do for SIL analyses (a sample SIL analysis is shown)Point out that that the FIELD DEVICES (red and yellow pie slices) contribute most to PFD (probability to fail dangerous) and Spurious trip rate.ASK: What should be improved to increase reliability? (answer = devices, diagnostics)
Slide 43 shows fault tree for analogue system with control in DCSMTTF = 15.9 yearsSlide 44 shows fault tree for FOUNDATION Fieldbus with Control in FieldMTTF = 48.2 yearsEquipment less prone to fail because of predictive intelligence and proactive intelligenceAbsolute values aren't realistic - as input comes from safety figures; but relative values are robustI’ve heard of projects where “all the critical loops are left 4-20 mA”. In light of this analysis, would you agree to such a strategy?
Shin-Etsu plant, the card in DCS (computer) failed, but instead of shutting down the plant, able to use Control in the Field for direct communication between measurement device and valve - continue operating while change card in DCS - avoid shutdownThis is the inherent back-up capability of CIF
Following findings based on study by Industrial Systems and Control Ltd - 'Control in the Field: Analysis of Performance Benefits'Industrial Systems and Control Ltd is spin-off from Strathclyde UniversityControl Engineering Consultancy, and TrainingStudy instigated by Fieldbus Foundation EMEA
Compare CIF:This example has a Rosemount 8800D FlowmeterFisher Valve Positioner20ms - AI execution in flowmeter30ms - data transfer from flowmeter to PID (proportional–integral–derivative controller) in valve positioner30ms - PID execution in valve positioner25ms - execution of instruction in valve positionerTotal latency - 105msWhich is lower than sample rate - 150ms (how often sample for instructions)With Control in Process Control System:For same set up but with control in process control system, not field20ms - AI execution in flowmeter30ms data transfer from flowmeter to PID in PCS20ms - PID execution in PCS30ms - data transfer from PCS to valve positioner25ms - execution of instruction in valve positionerTotal latency - 125msBut, asynchronous, so latency increases to 625ms because of jitter [ask for clarification on how this works if needed]Sample rate is 500msAssessment was made in simulationA simple continuous process model, coupled to a discrete PI controllerIMPORTANT - controllers tuned to same stability to allow comparisonRepeated for different process dynamicsSpeed of response to set point change and disturbance rejection assessedResults follow on next few slides
Only show Setpoint, case 1 - CIF, and case 3 - asyncDifference in performance is difference in time taken for process output to settle at 60% for 1 and 3
Only show Setpoint, case 1 - CIF, and case 3 - asyncDifference in performance is difference in time taken for process output to settle at 60% for 1 and 3
Only show Setpoint, case 1 - CIF, and case 3 - asyncDifference in performance is difference in time taken for process output to settle at 60% for 1 and 3
Only show Setpoint, case 1 - CIF, and case 3 - asyncDifference in performance is difference in time taken for process output to settle at 60% for 1 and 3
Only show Setpoint, case 1 - CIF, and case 3 - asyncGraph shows ability of process control to get process output to return to setpoint without deviating to greatly when disturbed[e.g. like ability of refrigerator to get back to 2 degrees without going too hot or cold when the door is opened a lot]CIF deviates far less than case 3, which ends up quite far from the set-pointFor fast process loops (e.g. flow, some temperature) CIF provides 40-60% faster settling time than asynchronous control in the DCS
“Control in the field: analysis of performance benefits” study from ISC (industrial systems and control)http://www.isc-ltd.com/
If control is not tight, need to leave large margin for error in case disturbance means values go beyond control limitBecause control is tighter with CIF, it's possible to set setpoint much nearer to the control limitCan push control limits harder - aiding e.g. energy efficiency / product quality /raw material utilisation
Great for theoretical and simulation, what about real world empirical tests?
The test bed was a liquid pressure control in the supplier’s flow lab – some decent sized valves and pipe.Using high-speed monitoring / recording, they measured the “response time” defined as the time from a detectable disturbance or change in the PV, until the air signal to the valve began to change from steady state.RED line is Control-in-fieldYELLOW line is pure 4-20 mA (on their host only)BLUE / GREEN is control in host with various ratios of controller cycle time / FF macrocycle time.DASHED line is response period = configured cycle timeCONTROL RESPONSE PERIOD defined as the time from the introduction of a disturbance (measurement deviation) to detecting a signal to the valve actuator (change in pneumatic signal) using high-speed sensors and recording / test equipment (NOT DCS TRENDS).
Like Shell Global Solutions says, when you do control in the controller, you consume a lot of the segment cycle time (macrocycle) UNNECESARILY. FOUNDATION™ Fieldbus is DESIGNED for control in field devices – It’s equally or more reliableIt’s equal or faster than 4-20 mAYou’re relying on the devices ANYHOW . . . When users and their consultants insist on old-school CIC, it’s like choosing an outstanding steak and cooking it WELL DONE . . . It takes too long and when you eat it, you wonder “what’s so great about this? Might as well just get some ole’ HART Hamburger!
I don’t know of anyone who actually would install 16 devices including 8 control valves (loops) on a single segment, but this graphic shows that with judiciously chosen instruments, even EIGHT simple loops can execute in less than 0.5 seconds.Q: What do I mean by “judiciously chosen instruments”?(Choose the ones that can go fast) “Cadillac” shown, but Fisker-Karma and Toyota are catching up.