Experiences with the Digital Twin of a Depropanizer Reboiler

Paper from the AFRC 2017 conference titled Experiences with the Digital Twin of a Depropanizer Reboiler

Wikipedia defines digital twins as "computerized companions of physical assets that can; be used for various purposes. Digital twins use data from sensors installed on physical; objects to represent their near real-time status, working condition or position." With this; paper we present the validation and integration of a digital twin into the continuous; operations of the refinery process heater serving as a reboiler for the bottoms of a; depropanizer tower at the HollyFrontier Refinery in Woods Cross, Utah.; This digital twinning technology was implemented in the refinery, debugged, tested,; hardened and calibrated throughout 2016. In February of 2017 the newly developed; instrument was commissioned and turned-over to refinery engineers and operators. The; integration of this instrument into refinery operations has resulted in:; i. increased refinery safety: by knowing the peak temperature on each coil of the reboiler,; a revision in operating strategy was implemented to minimize danger from tube rupture; in this alkylation unit.; ii. increased reboiler longevity: the local temperature and process fluid vapor-liquid; fractions throughout the coil run length made available by this instrument has made; possible a revision in the control strategy to optimize heat input to the alkylation column; while maintaining the operation strategy of a partial (vs. total) reboiler.; iii. insight into refinery operations: by trending the local information made available from; this continuous measurement, refinery engineers have been able to explain previously; unexplained refinery operations.

CRSim A F R C 2 0 1 7 Experiences with the Digital Twin of a Depropanizer Reboiler P. Smith: CRSim Inc. J. Thornock, S. Smith, M. Hradisky: The University of Utah D. Smith, P. Emett, M. Hardy: APCO Inc. K. Daines, B. Harris: HollyFrontier Corp. ABSTRACT Wikipedia defines digital twins as "computerized companions of physical assets that can be used for various purposes. Digital twins use data from sensors installed on physical objects to represent their near real-time status, working condition or position." With this paper we present the validation and integration of a digital twin into the continuous operations of the refinery process heater serving as a reboiler for the bottoms of a depropanizer tower at the HollyFrontier Refinery in Woods Cross, Utah. This digital twinning technology was implemented in the refinery, debugged, tested, hardened and calibrated throughout 2016. In February of 2017 the newly developed instrument was commissioned and turned-over to refinery engineers and operators. The integration of this instrument into refinery operations has resulted in: i. increased refinery safety: by knowing the peak temperature on each coil of the reboiler, a revision in operating strategy was implemented to minimize danger from tube rupture in this alkylation unit. ii. increased reboiler longevity: the local temperature and process fluid vapor-liquid fractions throughout the coil run length made available by this instrument has made possible a revision in the control strategy to optimize heat input to the alkylation column while maintaining the operation strategy of a partial (vs. total) reboiler. iii. insight into refinery operations: by trending the local information made available from this continuous measurement, refinery engineers have been able to explain previously unexplained refinery operations. 1 of 8 INTRODUCTION Wikipedia defines a digital twin as follows: "Digital twins refer to computerized companions of physical assets that can be used for various purposes. Digital twins use data from sensors installed on physical objects to represent their near real-time status, working condition or position. One example of digital twins can be the use of 3D modeling to create a digital companion for the physical object. It can be used to view the status of the actual physical object, which provides a way to project physical objects into the digital world. For example, when sensors collect data from a connected device, the sensor data can be used to update a "digital twin" copy of the device's state in real time.The term "device shadow" is also used for the concept of a digital twin. The digital twin is meant to be an up-to-date and accurate copy of the physical object's properties and states, including shape, position, gesture, status and motion."1 General Electric says this about the value of a digital twin: "A digital twin is a dynamic digital representation of an industrial asset, that enables companies to better understand and predict the performance of their machines and find new revenue streams, and change the way their business operates. 2 In March of this year Forbes Magazine quoted Thomas Kaiser, SAP Senior Vice President of IoT, as saying: "Digital twins are becoming a business imperative, covering the entire lifecycle of an asset or process and forming the foundation for connected products and services. Companies that fail to respond will be left behind."3 Digital twins are indirect measurement devices, or instrument models, albeit rather sophisticated ones. As such these instruments require validation and calibration. The validation process quantifies the degree of uncertainty of the output quantities of interest from the digital twin to the uncertainty in the input to the digital model (i.e. the sensor data). The calibration process reduces the uncertainty in the model parameters based on independently acquired information (including uncertainty) about the instrument model output. The instrument model is then run in near real time, coupled with modern data acquisition systems to continuously produce indirect output measurements. High performance computing and Bayesian methods combine to provide an opportunity to 1 https://en.wikipedia.org/wiki/Digital_Twins 2 https://www.ge.com/digital/industrial-internet/what-digital-twin 3 Bernard Marr, Forbes, MAR 6, 2017, What Is Digital Twin Technology - And Why Is It So Important? 2 of 8 efficiently use much more sophisticated instrument models than have ever been used before to produce indirect measurements of ever more difficult quantities of interest. Thus, the digital twin. In two companion papers 4,5 by the same authors, we described a project to develop and deploy a continuous, online digital twin of a refinery process heater serving as a reboiler for the bottoms of a depropanizer tower at the HollyFrontier Refinery in Woods Cross, Utah. The development, deployment, and utilization of this digital twin proceed through five phases: • Phase 1) Develop the instrument model for the reboiler; thus, producing a proof-ofconcept for the digital twin that focused on the tube skin temperature measurement as the quantity of interest. • Phase 2) Compare static skin and process fluid temperature measurements for HollyFrontier refinery; thus, calibrating and validating the digital twin for the quantities of interest. • Phase 3) Develop map of operating conditions, using historical HollyFrontier refinery data running offline; thus, testing the dynamic range of the digital twin. • Phase 4) Deploy the online dynamic digital twin at the refinery as part of the control system. • Phase 5) Learning from the digital twin, improving process operations. The two papers previously referenced described phases 1-4 (development and deployment of the digital twin). In this paper, after a descriptive review of the specific reboiler being instrumented, we discuss phase 5: learning from the digital twin how to improve process understanding and process operation. 4 Continuous Online Monitoring of Fireside Tube Skin Temperature in a Depropanizer Reboiler, American Flame Research Committee (AFRC) annual meeting, Salt Lake City, UT, September 2015. 5 Part II. Deployment of Continuous Online Monitoring of Local Tube Metal Temperature of a Refinery Depropanizer Reboiler, American Flame Research Committee (AFRC) annual meeting, Kauai, HI, September 2016. 3 of 8 THE REBOILER This fuel gas fired furnace acts as a reboiler for the bottom of the depropanizer tower. The furnace adds 25.6 MMBTU/hr to the bottom of this tower. It was designed for ~65 wt.% evaporation (91,750 lb/hr liquid, 176,000 l b / h r v a p o r ) o f c h a rg e at d e s i g n conditions with an efficiency of 85%. The design maximum pressure is 500 psig at 450oF temperature.  The reboiler furnace normally operates at 285 psig at 355oF at the inlet and 270 psig at 424-440oF at the outlet. This is a cylindrical, vertical type furnace a n d i s fi re d f ro m f o u r J o h n Z i n k PSFG-16RM staged fuel gas burners with individual gas pilots.  The burners are mounted in the floor of the furnace.  There are 120 tubes in the depropanizer reboiler heater.  There are 64 finned and 16 bare or shield, horizontal tubes in the convection section.  The shield tubes are located below the finned tubes to give a more even distribution of heat to the product and shield the finned tubes in the convection section.  The 40 radiant tubes are 24'-0" long except the radiant inlet tubes which are 27'-6" long and the Figure 1: Photograph and schematic cross section of the HollyFrontier refinery AlkyDepropanizer Reboiler showing the 4 coil locations. radiant outlet tubes which are 26'-6" long.  The radiant tubes are arranged vertically around the wall of the furnace. The reboiler heater has four identical coil passes. Each pass is equipped with 4 of 8 transmitters to show the individual coil flows, and exit process fluid temperature. Each coil pass has 16 finned convection tubes, 4 shield tubes, and 10 tubes in the radiant section.  All the tubes are 4.5" O.D. A-106 Gr. B stainless steel material.  All the return bends are welded.  The flow enters each coil pass on the top row of the convection section and flows down through the convection sections, out the cross-over and into the top of each radiant section.  Flow enters each radiant coil pass, flows through the 10 tubes, and out the top of the radiant section and returns to the depropanizer tower.  The radiant section northwest coil pass and north east coil pass outlets are on the north side of the furnace.  The radiant section south west coil pass and south east coil pass outlets are on the south side of the furnace. The furnace is 10'-5" diameter and 35'-0" tall, excluding the stack.  The stack is 4'-0" diameter and 82'-0" tall and rests upon a transition section and rectangular convection section on top of the cylindrical part of the furnace.  A damper is provided in the furnace stack and is manually operable from ground level.  Smothering steam lines are provided for the firebox.  The combustion air to the furnace is controlled by a damper system in the wind box on the bottom of the furnace.  A flue gas sample connection is provided in the stack below the damper for running flue gas analysis (for pollution control and efficiency). T H E D I G I TA L T W I N The digital twin as an instrument model developed for the continuous indirect measurements of the coil skin temperatures has been described previously6 . Validation of the digital twin occurs concurrently from 25 sensors providing continuous data from the physical twin. Thus, the SimMapp™ digital twin measurements are continuously validated through the online physical twin sensors data. These data, the validation process, the offline testing, the on-line deployment, as well as the calibration process were described in a companion paper7. 6 ibid., September 2015. 7 ibid., September 2016. 5 of 8 Two major sources of uncertainty contribute to measurement error in the digital twin: 1. instrument model parameter uncertainty, 2. and measurement uncertainty in the 25 measured inputs to the instrument model. Both of these sources of uncertainty are propagated through the instrument modeling to produce a nonlinear response in the output of the digital twin. This uncertainty in the predicted quantities of interest is then reported as part of the continuous twinned output. This approach offers the opportunity to use Bayesian methods to learn from the data to reduce the uncertainty in the quantities of interest by narrowing the uncertainty in the model parameters and in the 25 measured inputs. This learning algorithm allows the uncertainty to evolve as a function of the history of the digital and physical pair. This also constitutes a validation process for the digital twin. LEARNING FROM THE TWO TWINS This SimMapp™ digital twinning technology was implemented in the refinery, debugged, tested, hardened and calibrated throughout 2016. In February of 2017 the newly developed instrument was commissioned and turned-over to refinery engineers and operators. The pairing of the digital and physical twins allowed for dynamic learning about the process and the process equipment that was not obtainable from sensor data alone. The source of this learning was the information available from the digital twin. The sensor data from the physical twin authenticated the quantities of interest from the associated digital twin. As the digital twin became more accepted by the engineers and operators, more information was extracted from it. While data mining is still ongoing, several lessons have been learned already: 1. The digital twin was initial commissioned to provide tube skin temperature for the entire length of each of the four coils. While the validation process described above ensures that the instrument model is consistent with the dcs data collected by the refinery for the reboiler, it does not uniquely define the tube metal skin temperatures. The heat transfer coefficient for each coil requires independent information. Initially this information was calibrated from periodic FLIR data obtained by the refinery. This 6 of 8 calibration process was described in a previous paper8. After commissioning the digital twin, the refinery activated two thermocouples along each of the four coils. These continuous measurements are shown as red circles in Figure 2. These additional data from the physical twin corroborated the twinning process including the uncertainty analysis. It gave the refinery personnel confidence in other data coming from SimMapp™ and served as the basis for decision making on other quantities of interest. Figure 2: Tube metal skin temperature (solid red line) as a function of tube length from inlet to outlet of the southwest coil in the reboiler. The process fluid temperature (solid green line) and the vapor fraction (quality) of the process fluid (solid blue line) are also shown. The dotted lines around each solid line shows the uncertainty in the measurements at this point in the learning algorithm. The green circle shows the measurement of the process fluid outlet temperature as reported by the refinery dcs. The dcs report of the tube skin temperature as measured by two thermocouple measurements installed on this coil are shown with the red circles. The message across the bottom of this figure indicates learned reasons why the twins are reporting an uncertain range that is larger than expected from past their historical experience. The maximum skin temperature, its uncertainty and the timestamp for this coil are all listed in the figure title. 8 ibid., September 2016. 7 of 8 2. The SimMapp™ digital twin has increased refinery safety. By knowing the peak temperature on each coil of the reboiler, a revision in operating strategy was implemented to minimize danger from tube rupture in this alkylation unit. It had long been assumed that much was being missed by observing only two temperatures on each coil. The data from the digital twin bore out this thesis and provides the missing information for continuous operation at or below the critical tube metal operating temperature. 3. The SimMapp™ digital twin has increased reboiler longevity. While the initial quantity of interest for implementing the digital twin was to track the local tube metal temperature, it became apparent that the information available from SimMapp™ about the local process fluid vapor-liquid fractions throughout the coil run length was valuable to decrease fouling and understand operability. The local vapor fraction made available by this instrument (as shown in Figure 2) has made possible a revision in the control strategy to optimize heat input to the alkylation column while maintaining the operation strategy of a partial (vs. total) reboiler. 4. The SimMapp™ digital twin has provided insight into refinery operations. By trending the local information made available from this continuous measurement, refinery engineers have been able to explain previously unexplained refinery operations. CONCLUSIONS We have developed and deployed a digital twin for the the HollyFrontier Woods Cross refinery alky-depropanizer reboiler. This new continuous measurement, monitoring and prediction tool (called SimMapp™) links high performance computing (hpc) simulation methods with combustion computational fluid dynamic (CFD) large eddy simulations (LES) and Bayesian analysis methods together with sensor data from the physical appliance. Calibration, validation, offline testing, and online deployment of this technology was completed in the first quarter of 2017. Utilization of this digital twinning technology has helped engineers and operators, among other things, to continuously track local tube metal temperature and process fluid vapor fraction in order to: (i) increase refinery safety, (ii) increase reboiler longevity, and (iii) increase their insight into refinery operations. 8 of 8