[en] In this paper, we report on the experiment carried out at the Frascati Tokamak Upgrade (FTU) on the porting of the plasma control system (PCS) from a LynxOS architecture to an open source Linux real-time architecture. The old LynxOS system was implemented on a VME/PPC604r embedded controller guaranteeing successful plasma position, density and current control. The new RTAI-Linux operating system has shown to easily adapt to the VME hardware via a VME/INTELx86 embedded controller. The advantages of the new solution versus the old one are not limited to the reduced cost of the new architecture (based on the open-source characteristic of the RTAI architecture) but also enhanced by the response time of the real-time system which, also through an optimization of the real-time code, has been reduced from 150 μs (LynxOS) to 70 μs (RTAI). The new real-time operating system is also shown to be suitable for new extended control activities, whose implementation is also possible based on the reduced duty cycle duration, which leaves space for the real-time implementation of nonlinear control laws. We report here on recent experiments related to the optimization of the coupling between additional radiofrequency power and plasma

[en] The ITER CODAC (COntrol, Data Access and Communication) conceptual design resulted from 2 years of activity. One result was a proposed functional partitioning of CODAC into different CODAC Systems, each of them partitioned into other CODAC Systems. Considering the large size of this project, simple use of human language assisted by figures would certainly be ineffective in creating an unambiguous description of all interactions and all relations between these Systems. Moreover, the underlying design is resident in the mind of the designers, who must consider all possible situations that could happen to each system. There is therefore a need to model the whole of CODAC with a clear and preferably graphical method, which allows the designers to verify the correctness and the consistency of their project. The aim of this paper is to describe the work started on ITER CODAC modeling using Matlab/Simulink. The main feature of this tool is the possibility of having a simple, graphical, intuitive representation of a complex system and ultimately to run a numerical simulation of it. Using Matlab/Simulink, each CODAC System was represented in a graphical and intuitive form with its relations and interactions through the definition of a small number of simple rules. In a Simulink diagram, each system was represented as a 'black box', both containing, and connected to, a number of other systems. In this way it is possible to move vertically between systems on different levels, to show the relation of membership, or horizontally to analyse the information exchange between systems at the same level. This process can be iterated, starting from a global diagram, in which only CODAC appears with the Plant Systems and the external sites, and going deeper down to the mathematical model of each CODAC system. The Matlab/Simulink features for simulating the whole top diagram encourage us to develop the idea of completing the functionalities of all systems in order to finally have a full simulation of ITER CODAC

[en] Highlights: ► New magnetic shape control system has been implemented. ► It has been intensively tested in a simulation environment. ► A tool chain to produce LTI model and simulate its behaviour has been implemented. ► Experimental results are shown. -- Abstract: The Mega Ampere Spherical Tokamak (MAST) real time plasma position controller is based on an optical linear camera placed on the mid plane of the vessel. This solution has the advantage of being a direct observation of the D_α emissions coming from the interaction between the boundary of the plasma and neutral gas, but, on the other hand, it restricts the control to the outer radius of the plasma only. A complete chain of tools has been set up to implement, test and simulate a new real time magnetic plasma shape controller based on the rtEFIT code. The complete working path consists of three elements: a linear static relationship between control parameters and current demands, a linear state space model needed to represent the plasma dynamic response in closed loop simulations, and the possibility to run simulations inside the Plasma Control System (PCS). The linear relationship has been calculated using the FIESTA code, which is developed using Matlab at CCFE. The linear state space model was generated using the CREATE-L code developed by the CREATE Consortium. It has already been successfully used to model JET, FTU and TCV tokamaks. Using this working path many simulations have been carried out allowing fine tuning of the control gains before the real experiment. The simulation testing includes the plasma shape control law as implemented in PCS itself, so intensive debugging has been possible prior to operation. Successful control using rtEFIT was established in the second dedicated experiment during the MAST 2011–12 campaign. This work is a stepping stone towards divertor control which is ultimately intended for application to the super-X divertor in the MAST Upgrade experiment

[en] The Feedback Control System running at FTU has been recently ported from a commercial platform (O.S. LynxOS) to an open-source GNU/Linux-RTAI platform, obtaining significant performance and cost improvements. Thanks to the new platform, more user friendly tools can be developed in order to help the designer with new control laws. A relevant goal within this new framework is to provide a high level environment where new control algorithms can be created then simulated and finally released without minding the code implementation issues. The ideal situation would be to have a dedicated framework which provides all the necessary phases from the design to the commissioning of the new software. This framework should simulate the real-time context and make transparent to the user on the one hand all the issues related to the simulation (e.g. experimental data retrieving) and on the other hand all the aspects (platform, operating system, programming language, network, hardware...) related to the actual environment where the new algorithm will be run. In this paper we report on recent developments, based on The MathWorks' Simulink and Real Time Workshop (RTW) packages, aimed at obtaining the above mentioned environment where a new control law can be easily modelled, simulated with the real time constraints and then translated in the appropriate executable format. Using this tool, the control designer only needs to specify the control law in the Simulink graphical environment. The arising model is then automatically translated in C code, integrated with control system code and simulated in real-time using the data from the FTU data base archive. All the necessary steps to adapt the RTW scripts and the control system code to the new simulation/validation environment will be illustrated in this paper. Moreover we will report on some experimental tests where the actual experiment is compared with the simulations provided by the proposed environment. (author)

[en] Highlights: ► We applied linearized models for a new magnetic control on MAST tokamak. ► A suite of procedures, conceived to be machine independent, have been used. ► We carried out model-based simulations, taking into account eddy currents effects. ► Comparison with the EFIT flux maps and the experimental magnetic signals are shown. ► A current driven model for the dynamic simulations of the experimental data have been performed. -- Abstract: The aim of this work is to provide reliable linearized models for the design and assessment of a new magnetic control system for MAST (Mega Ampère Spherical Tokamak) using rtEFIT, which can easily be exported to MAST Upgrade. Linearized models for magnetic control have been obtained using the 2D axisymmetric finite element code CREATE L. MAST linearized models include equivalent 2D axisymmetric schematization of poloidal field (PF) coils, vacuum vessel, and other conducting structures. A plasmaless and a double null configuration have been chosen as benchmark cases for the comparison with experimental data and EFIT reconstructions. Good agreement has been found with the EFIT flux map and the experimental signals coming from magnetic probes with only few mismatches probably due to broken sensors. A suite of procedures (equipped with a user friendly interface to be run even remotely) to provide linearized models for magnetic control is now available on the MAST linux machines. A new current driven model has been used to obtain a state space model having the PF coil currents as inputs. Dynamic simulations of experimental data have been carried out using linearized models, including modelling of the effects of the passive structures, showing a fair agreement. The modelling activity has been useful also to reproduce accurately the interaction between plasma current and radial position control loops

[en] In this paper we report on recent developments, based on The MathWorks' Simulink and Real-Time Workshop (RTW) packages, aimed at obtaining a high level environment where a new control law can be easily modelled, simulated with the real-time constraints and then translated in the appropriate executable format. Using this tool, the control designer only needs to specify the control law in the Simulink graphical environment. The arising model is then automatically translated in C code, integrated with control system code and simulated in real-time using the data from the Frascati Tokamak Upgrade (FTU) data base archive. All the necessary steps to adapt the RTW scripts and the control system code to the new simulation/validation environment will be illustrated in this paper. Moreover, we will report on some experimental tests where the actual experiment is compared with the simulations provided by the proposed environment

[en] The next nuclear fusion experiment, ITER, is providing the infrastructure for the optimal operation of a burning plasma, requiring feedback control of discharge parameters and on-line evaluation of computationally intensive models running in a cluster of controller nodes. Thus, the synchronization of the available information on the plasma and plant state variables among the controller nodes is a key issue for ITER. The ITER conceptual design aims to perform feedback control on a cluster of distributed controllers connected by a Synchronous Databus Network (SDN). Therefore it is mandatory to achieve a deterministic data exchange among the controller nodes with a refresh rate of at least 1 kHz and a jitter of at least 50 μs. Thus, a conservative estimate of the data flow within the controller network can be 3 kSample/ms. In this paper the open source RTnet project is evaluated to meet the requirements of the SDN of ITER. A testbed involving a cluster of eight nodes connected over a standard ethernet network has been set up to simulate a distributed real-time control system. The main goal of the test is to verify the compliance of the performance with the ITER SDN requirements

[en] The Feedback Control System running at FTU has been recently ported from a commercial platform (based on LynxOS) to an open-source GNU/Linux-based RTAI-LXRT platform, thereby, obtaining significant performance and cost improvements. Based on the new open-source platform, it is now possible to experiment novel control strategies aimed at improving the robustness and accuracy of the feedback control. Nevertheless, the implementation of control ideas still requires a great deal of coding of the control algorithms that, if carried out manually, may be prone to coding errors, therefore time consuming both in the development phase and in the subsequent validation tests consisting of dedicated experiments carried out on FTU. In this paper, we report on recent developments based on Mathworks' Simulink and Real Time Workshop (RTW) packages to obtain a user-friendly environment where the real time code implementing novel control algorithms can be easily generated, tested and validated. Thanks to this new tool, the control designer only needs to specify the block diagram of the control task (namely, a high level and functional description of the new algorithm under consideration) and the corresponding real time code generation and testing is completely automated without any need of dedicated experiments. In the paper, the necessary work carried out to adapt the Real Time Workshop to our RTAI-LXRT context will be illustrated. A necessary re-organization of the previous real time software, aimed at incorporating the code coming from the adapted RTW, will also be discussed. Moreover, we will report on a performance comparison between the code obtained using the automated RTW-based procedure and the hand-written C code, appropriately optimised; at the moment, a preliminary performance comparison consisting of dummy algorithms has shown that the code automatically generated from RTW is faster (about 30% up) than the manually written one. This preliminary result combined with the fact that the use of RTW eliminates the coding workload leads to the conclusion that this approach to control implementation grants significant advantages. According to the FTU experiment program, the actual deployment of the new tool is scheduled for the next experimental campaign

[en] In this paper, we will report on the experimental results arising from the implementation of optimization techniques to maximize the RF power coupling versus the plasma conditions in the FTU experimental facility. These experiments are carried out by employing the open-source Linux-RTAI control system currently running on the FTU digital feedback loop. The RF power source under consideration is a lower hybrid system (LH) based on six gyrotrons with a nominal power output capability of 1.1 MW each. The optimization of the coupling level between the plasma and the emitting antenna reduces the reflected power, thus maximizing the heating effects in addition to avoiding danger to the emitter (equivalently, annoying safety shutdowns of the system). To this aim, the plasma displacement is modified by suitably adjusting the reference input to the stabilizing feedback, according to a steepest descent algorithm. It will be shown in the paper how this algorithm achieves a satisfactory level of robustness with respect to measurement errors and well performs both in simulation and in experimental tests, thus leading to an improved effectiveness of the RF heating system

[en] In this paper we focus on the instabilities caused by the nonlinear behavior of the F coil current amplifier at FTU. This behavior induces closed-loop instability of the horizontal position stabilizing loop whenever the requested current is below the circulating current level. In the paper we first illustrate a modeling phase where nonlinear dynamics are derived and identified to reproduce the open-loop responses measured by the F coil current amplifier. The derived model is shown to successfully reproduce the experimental behavior by direct comparison with experimental data. Based on this dynamic model, we then reproduce the closed-loop scenario of the experiment and show that the proposed nonlinear model successfully reproduces the nonlinear instabilities experienced in the experimental sessions. Given the simulation setup, we next propose a nonlinear control solution to this instability problem. The proposed solution is shown to recover stability in closed-loop simulations. Experimental tests are scheduled for the next experimental campaign after the FTU restart.