CEC Conference, July 10, 2017 Comparison of different cryogenic control strategies via simulation applied to a superconducting magnet test bench at CERN.

Slides:



Advertisements
Similar presentations
ERT 210 Process Control & dynamics
Advertisements

استاد محترم : دکتر توحيدخواه ارائه دهنده : فاطمه جهانگيري.
Automation I. Introduction. transmitter actuator Structure of control system Process or plant Material flow sensorstransducers actuating units actuating.
Control Architectures: Feed Forward, Feedback, Ratio, and Cascade
Process Control: Designing Process and Control Systems for Dynamic Performance Chapter 6. Empirical Model Identification Copyright © Thomas Marlin 2013.
1 Finding good models for model-based control and optimization Paul Van den Hof Okko Bosgra Delft Center for Systems and Control 17 July 2007 Delft Center.
PROCESS INTEGRATED DESIGN WITHIN A MODEL PREDICTIVE CONTROL FRAMEWORK Mario Francisco, Pastora Vega, Omar Pérez University of Salamanca – Spain University.
280 SYSTEM IDENTIFICATION The System Identification Problem is to estimate a model of a system based on input-output data. Basic Configuration continuous.
INTEGRATED DESIGN OF WASTEWATER TREATMENT PROCESSES USING MODEL PREDICTIVE CONTROL Mario Francisco, Pastora Vega University of Salamanca – Spain European.
NORM BASED APPROACHES FOR AUTOMATIC TUNING OF MODEL BASED PREDICTIVE CONTROL Pastora Vega, Mario Francisco, Eladio Sanz University of Salamanca – Spain.
The Mechatronics Design Lab Course at the University of Calgary Presented June 2, 2003.
Get More from your Process using Modelling and Simulation Andy Clegg Industrial Systems and Control Ltd.
CSE 425: Industrial Process Control 1. About the course Lect.TuLabTotal Semester work 80Final 125Total Grading Scheme Course webpage:
Proportional/Integral/Derivative Control
CIS 540 Principles of Embedded Computation Spring Instructor: Rajeev Alur
Static Pressure Control Loop The purpose of the static pressure control loop is to maintain an optimal static pressure in the ductwork. The control loop.
2 Technical seminar - 5 th June 2014 Vincent Roger - TE-MSC-TF.
HQ TEST CERN by Marta Bajko CERN TE-MSC TF For Hi Lumi and LARP the 16 th of November 2011 CERN.
21/01/02 - ECAL Cooling - Arnaud Hormiere ST/CV 1 Development of ECAL COOLING PLANT Application to a Super Module.
Book Adaptive control -astrom and witten mark
LHC ARC Commissioning report during LS1 Agenda: VRGPE documentation (former VRJGE) Active Penning modification By-Pass Valves modification LHC ARC commissioning.
Multiple Model approach to Multi-Parametric Model Predictive Control of a Nonlinear Process a simulation case study Boštjan Pregelj, Samo Gerkšič Jožef.
MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator ARG Rong Zhang MIMO LQG/LTR Control for the Earthmoving Vehicle Powertrain Simulator.
Model Reference Adaptive Control (MRAC). MRAS The Model-Reference Adaptive system (MRAS) was originally proposed to solve a problem in which the performance.
Low Level Control. Control System Components The main components of a control system are The plant, or the process that is being controlled The controller,
Modeling and simulation of cryogenic processes using EcosimPro
ERT 210/4 Process Control Hairul Nazirah bt Abdul Halim Office: CHAPTER 8 Feedback.
Lecture 25: Implementation Complicating factors Control design without a model Implementation of control algorithms ME 431, Lecture 25.
L. Serio COPING WITH TRANSIENTS L. SERIO CERN, Geneva (Switzerland)
Matlab Tutorial for State Space Analysis and System Identification
Process Definition of the Operation Modes for Super-FRS Magnet Testing CSCY - CrYogenic department in Common System, GSI, Darmstadt Y. Xiang, F. Wamers.
Plan for test station Marta Bajko For the Technical Review of FReSCa2 June 2015 Saclay Paris.
AT-ACR B. VULLIERMECSOC Meeting 29 September Cryogenics for LHC Test Benches Safety Aspects Overview of the Test Station Overview of the Operation.
HIE ISOLDE commissioning of the cryogenic control system installation at CERN. LNL-INFN 22/07/2015 Dr. Marco Pezzetti (CERN )
Cascade Control Systems (串级控制系统)
CIS 540 Principles of Embedded Computation Spring Instructor: Rajeev Alur
Cryogenics Operations 2008, CERN, Geneva, Switzerland 1 CRYOGENICS OPERATIONS 2008 Organized by CERN Industrial contributions to the maintenance of CERN.
Chapter 20 Model Predictive Control (MPC) from Seborg, Edgar, Mellichamp, Process Dynamics and Control, 2nd Ed 1 rev. 2.1 of May 4, 2016.
SOFTWARE: A SOLUTION FOR MODEL PREDICTIVE CONTROL SOFTWARE: A SOLUTION FOR MODEL PREDICTIVE CONTROL Laboratoire.
Mechatronics at the University of Calgary: Concepts and Applications
R. Panchal, GLN Srikanth, K. Patel, P. Shah, P. Panchal,
Process Dynamics and Operations Group (DYN) TU-Dortmund
PROMONICON Software for Process Monitoring, Operation and Control
Automatic Control Theory
A Microcontroller Based Power Management System for Standalone Micro grids With Hybrid Power Supply 1.
HL-LHC IT STRING and Series test of SC link
Presentation CHEMCAD Simulation Software Training Simulator
Cryogenic behavior of the cryogenic system
IPC (GROUP-5) SR. NO NAME EN. NO 1 SAVAN PADARIYA
Presentation at NI Day April 2010 Lillestrøm, Norway
PID tuning & UNICOS PID auto-tuning
Innovative He cycle Francois Millet.
Dana M. Arenius Jefferson Laboratory Cryogenics Dept Head
SPS cryogenic proximity equipment and SM18 validation
Dynamic simulations: a useful tool for cryogenic installations
Hollow e- lens, Cryogenic aspects
CRYOGENICS OPERATIONS 2008 Organized by CERN
MPC in Statoil Stig Strand, specialist MPC Statoil Research Center 93  SINTEF Automatic Control Dr. ing 1991: Dynamic Optimisation in State.
CRYOGENICS OPERATIONS 2008 Organized by CERN
Block 4 and Cluster D - Safety
Overall Objectives of Model Predictive Control
Linear Control Systems
SHMS Cryogenics and Q2(Q3Dipole) Cool Down
LINAC Heat Management A New Approach
Cooling aspects for Nb3Sn Inner Triplet quadrupoles and D1
Process Control Engineering
Model-based Predictive Control (MPC)
CERN with UNICOS M. Pezzetti AT_ECR_CE 27 June 2005.
PID Controller Design and
Cryogenic management of the LHC Run 2 dynamic heat loads
Presentation transcript:

CEC Conference, July 10, 2017 Comparison of different cryogenic control strategies via simulation applied to a superconducting magnet test bench at CERN M. PEZZETTI, Prof. Pasquale ARPAIA, Prof. Hervé COPPIER, Prof. Mario DI BERNARDO, Donato DE PAOLA, Agostino GUARINO, Benjamin LUZ PEDEMONTE CERN, 1211 Geneva 23, Switzerland Aknowledegments: the cryo operator team

INDEX Introduction Model synthesis and analysis Proposal Hi Lumi – LHC HFM Schema HFM Pre - cooling Model synthesis and analysis ECOSIM Switched Control Schema Cascade Control Schema Proposal Advanced Control Adaptive Cascade Controller Model Predictive Controller Controllers Comparison Conclusions structured

HFM Schema Biggest CRYOSTAT present at CERN

HFM Pre-cooling Pre – Cooling process: Supply line Actuators 300K 80K The way to reach the steady state temperature of 1.9 K is divided in three parts: Precooling from 300 K to 80 K, lead by GHe [10h – 200h, adjustable for each magnet]; Cooling from 80 K to 4.5K, lead by LHe [10h]; Cooling from 4.5 K to 1.9 K, lead pumping superfluid LHe [24h]. Pre – Cooling process: Supply line Actuators TWO VALVES Physical Constraints: Pipe Pressure < 3.5 - 4 bar Max Flow 50 to 90 g/s Difference of temperature on the magnet 50 K or lower.

Switched Control Schema Cold Valve(CV870) Warm Valve (CV871) PID 1 PID 4 PID 2 PID 5 PID 3 PID 6 3 different PIDs on each valve. The control signal is the minimum between PIDs outputs PRO: Intuitive Schema, a PID for each constraint for each valve. CONS : semi-automatic control need an intensive operation work!

Cascade Control Schema Current Controller PRO: Well organised strategy. PIDs cascade follows the logical steps of flow control with a smooth mix of cold and warm gaseous helium. CONS: Satisfaction of each constraints at the same time is guaranteed only in presence of nominal conditions. When this does not happens, some parameters exceed their limits.

EcosimPro EXPERIMENT SCHEMATIC Model tuned to fit real data taken from the field. EXPERIMENT Parameters Solver Tolerance SCHEMATIC EcosimPro is a Modelling and Simulation Software. Availability of a Cryogenic Library (CRYOLIB) developed at CERN. Chance to validate control schemas in silico. SIMULATIONS

Identification Strategy CV870 FT870 CV871 FT871 PT874 TT809a TT809b TT809d TT809e Our proposal is to identify a black box model of the plant in order to develop: A new Adaptive Cascade Controller A Model Predictive Controller Moreover, the controllers have to cool down the system in the most uniform way possible (ideally, following a ramp) even in presence of disturbances. The controllers must be user-friendly for the operators. TT800g At 80 K At 300 K

Adaptative Cascade Controller proposal Reference: Cool Down Nominal Time Constraints: Delta T Pressure Flow Pro: Uniform Cooling User Friendly Simple to implement Cons: (Chattering)

Inner Loop It works on the basis of a temperature error. Piping System Model is a Discrete Time System. (Discrete) PI tuned empirically (MATLAB tool). The loop is able to follow a descending ramp reference with no error. Spike on the Warm Valve opening due to the exceeding of physical limit. Temperature

External Loop Controller Constraint controllers give as output a signal between 0 and 1. Different strategies to choose the way they evaluate this output. Currently, the simplest choice has been tested. Performance oriented and Safety oriented solution could be tested.

Performances Simulation Details: Nominal Time: 7 Hours Delta T: 30K Flow: 90 g/s

MPC Proposal Uses the Piping System Model to calculate the present and anticipate n-step ahead future responses by modelling a prediction model. Optimizes the valves openings adopting a linearly constrained quadratic programming. It converges to the desired temperature set point by minimizing its tracking error, rate-of-change of the inputs.

MPC Block Schema

MPC Simulation Linear Discrete Time Piping System Model Temperature set point fixed from 300 K to 80 K. Optimizer block generates reference tracking error based on the prediction horizon.

Comparison results of the control models 1/2 Cryogenics parameters: 8 Hours, 30 K delta T, flow, 60 g/s, 2,5 bar pressure*. Working with the same operative conditions, ACC presents a faster response while MPC provides a smoother one. Choosing a different strategy for the ACC external controller, the performances could be enhanced, smoothing the response and reducing oscillations.

Comparison 2/2 Regarding Delta T, Figure 14 a, both controllers satisfy the constraint. ACC approaches the limit faster than MPC, but its response presents more oscillations than the other MPC evaluates the flow on the basis of a linearly constrained quadratic programming, increasing the cold flow and decreasing the warm one linearly. ACC has a different mixing strategy, in which the total flow decreases during time. Because of the strategy chosen in the external controller, oscillations arise when the Delta T approaches to its constraint.

Conclusions This study enhance the possibility to use advance control (user friendly!) in real installation at CERN. The difference between the proposed controllers is that ACC provides a faster response with some oscillations, while MPC is slower but smoother, both method will not need an operation supervision during cold down phase. ACC solution is more user friendly than MPC and it’s easier to implement but more difficult to use in a PLC environment.

Characteristic of the Valves and Pressure Fitting Electrical – Hydraulic Analogy

Piping system identification Open loop cycle Fit to Estimation Real Data Y= 57.16% Matlab identification toolbox Transfert function equation Closed loop Cycle Fit to Validation data: 81.22% Good dynamic replication RMS: 9.93K Estimation Method: Process Model 𝐺 𝑢 = 𝐾 𝑢 𝑒 − 𝑇 𝑑𝑢 𝑠 (1+ 𝑇 𝑝1𝑢 𝑠)(1+ 𝑇 𝑝2𝑢 𝑠) u=1,2

Cryostat Identification Fit to Estimation Real Data Y1= 98.52% Y2= 96.03% Y3= 98.28% Y4= 86.47% Fit to Data: 63.17% RMS: 11.09K Fit to Data: 50.45% RMS: 20.24K Estimation Method: Subspace Method for State Space Identification 4th order dx/dt=Ax+Bu y=Cx Fit to Data: 62.52% RMS: 11.33K Fit to Data: 57.26% RMS: 17.43K Good dynamic replication