Control Sheet No. 16


Korea is definitely the most "acceleratized" Country

By: M. Pleško (COSYLAB)

We are all impressed by the economic growth of Korea, and we're even more impressed at the commitment of Koreans at the national level to participate in "Big Physics" research. For one, even though Korea is by far smaller than other ITER member countries, it participates as equal partner and even more, it has a clear strategy path to a full fusion-based electric plant [1-3]. Their accelerator activities are even more impressive. I have been closely following Korean accelerator activities since 1991, when a colleague from the Pohang Light Source (PLS), Antonio Choi, worked together with me on machine physics applications for a year during his visit to Elettra, where I was employed at that time. I am particularly proud that I have visited, in recent years, all the latest big projects under construction: PAL-XFEL [4], PEFP [5], RISP [6] and KHIMA [7], not to mention the PLS upgrade to PLS-II, which was finished in record time [8].

But this article is not about the scientific projects, which are constructed far away from the public eyes. Far from it, because with only such projects, even Korea couldn't qualify for the most "acceleratized" country, given that other countries also have many such projects. No, a country can be considered truly acceleratized only if accelerators are available to citizens on a large scale. And if you can't buy your 100 MeV linac at your local Radio Shack yet, such accelerators should at least be available in public places.


Which is exactly the case of the linac in the photo from a street in downtown Seoul. OK, the particle source is missing, maybe also the klystron, whereas cabling may have been solved with a clever wireless solution. Probably most of the people passing by don't seem to think it is an accelerator, but this just shows how natural such a structure fits into the environment. Being who I am, now all I have to figure out is whether they would buy a control system for it :-)

But we have to be quick, because Koreans are not only strong in accelerators, but also in areas which compete with Cosylab. Well, at least Cosy, if not lab. See for yourselves the nice coffee bar that I found not far from the accelerator street. Or is this just a secret spin-off from Cosylab that uses the coffee bar as a front for its mission? We'll never know....












Predictive Diagnostics for High-Availability Accelerators

By: M. Gašperin (University of West Bohemia, Pilsen, Czech Republic), K. Žagar (COBIK*), D. Bokal (Cosylab), K. Strniša (Cosylab), G. Pajor (COBIK*), L. Medeiros-Romao, D. Vandeplassche (SKC-CEN, Mol, Belgium)

High availability of an accelerator is one of the key requirements in an Accelerator Driven System such as the proposed MYRRHA transmuter [1]. Accelerators consist of several crucial components and subsystems which are subjected to wear, material stress and environmental influences. These eventually cause the equipment to fail and can result in beam trips or even emergency shutdowns of an accelerator. Furthermore, if such a fault is unexpected, maintenance procedures can take longer than necessary [2].

Availability can be improved by either improving the reliability of individual components (increasing the Mean Time Between Failures (MTBF)) or by switching from preventive or reactive maintenance strategies to condition-based maintenance, which ensures minimum system downtime in case of breakdowns and faults (Mean Time To Recovery (MTTR)).


Predictive diagnostics strives to predict where a failure is likely to occur, so that mitigating actions can be taken in a more controlled manner. The aim is to prevent failure of other components while pinpointing exactly which component is about to fail. The impending fault can be identified by observing trends in the values of process variables that quantify inputs and outputs of components. In addition, sensors measuring vibration, temperature or noise can be attached to critical components.

Research shows that usually failures go through a distinct incipient phase [3]. This means that there are some noticeable indicators, which provide advanced warning about the onset of a failure. The role of automated Condition Monitoring (CM) is to timely detect this onset, localize the root-cause and, possibly, trend its progression over time. The remaining time until final breakdown can be long enough to allow for efficient maintenance service [4].

The basic stages of CM are feature extraction, feature evaluation and fault isolation (Figure 1). A feature is a function of the measured signal and should be sensitive only to the fault while insensitive to the operating conditions [5]. An important feature of condition monitoring systems is the prediction of the future evolution of the fault.



Figure 1: The PHM System

A Predictive diagnostics and Health Management framework (PHM) system is designed around the three main tasks: observation, analysis and action. The observation part consists of appropriate hardware and software components for signal acquisition and processing in order to compute the required feature values. The features are then analyzed for the presence of faults and possible trends. Finally, the high level decision support system assesses the probability of specific faults, remaining useful life and proposes the appropriate action.


Model-based fault detection relies on mathematical models to estimate the current condition of the system. The model can be built on physical equations (white- or grey-box model) or data (black-box model). The basis for fault detection and identification are the residual values, which are computed from comparing the model output to measured values, or comparing nominal and estimated system parameter values.

Fault identification and localization is performed from the Fault Signature Matrix (FSM), which connects faults and specific residuals (symptoms). Given a set of symptoms and a set of considered faults, the FSM encodes the relationship between the effects of a fault and each symptom. Fault isolation then consists of implementing an appropriate artificial intelligence method to search for the closest matching fault for the observed signature pattern (Figure 2).


Figure 2:  Model-based residual generation

PHM Methods for Mechanical Systems

Mechanical components are components in assemblies for cooling and vacuum systems and consist of pumps, blowers, bearings and gearboxes. The main sources of diagnostic information for mechanical systems are vibration signals [3], acoustic emissions and temperature. Symptoms for fault detection and isolation include time domain methods (RMS, variance, kurtosis, crest factor and time synchronous averaging), frequency domain methods (FFT spectra and power spectra) and time-frequency domain analysis (discrete and continuous wavelet analysis).

PHM Methods for Electronic Systems

Electronic systems include power supplies, electric drives in blowers and pumps, electronic systems, klystrons, etc. [7]. Systems for early fault detection and failure prediction continuously monitor current, voltage, and temperature signals. Along with sensor information, soft performance parameters such as loads, throughputs, queue lengths, and bit error rates can be tracked [8].


Prognostics deals with estimation of the Remaining Useful Life (RUL), i.e. the amount of time a component can be expected to continue operating within its given specifications [6]. The physics-based models rely on detailed physical modeling by means of the finite element method, which serves to compute spatial distributions of stresses in the material and their effect on the component health. The idea of data-driven methods is to make use of condition monitoring data to build the model and then use the model to predict a future trend (Figure 3).


Figure 3: Example of prognostics output


The vector of all feature values defines the state of the accelerator, and the predicted values of features define the trend of this state. These vectors must lie within the region of safe operation in the space of all possible accelerator states. The purpose of the PHM system is to detect when the state of the accelerator is moving towards the boundaries of the region of safe operation and either take automatic corrective action, or trigger an appropriate alarm for the operator to take action.


PHM is a promising technology that can be used within the maintenance decision-making process to provide failure predictions, increase the operational availability of systems, lower sustainment costs by reducing the costs and duration of downtime, improve inspection and inventory management, and lengthen the intervals between maintenance actions.

Financial support for the work was provided by the project EXLIZ – CZ.1.07/2.3.00/30.0013, which is co-financed by the European Social Fund and the state budget of the Czech Republic, and by COBIK*

*COBIK — Centre of Excellence for Biosensors, Instrumentation and Process Control (COBIK), Velika pot 22, SI-5250 Solkan, Slovenia COBIK is financed by the European Union, European Regional Development Fund and Republic of Slovenia, Ministry of Higher Education, Science and Technology.”


[1] D. Vandeplassche et al., “The MYRRHA Linear Accelerator” Proceedings of IPAC2011, San Sebastián, Spain, 2011

[2] H. Takei et al., “Estimation of Acceptable Beam Trip Frequencies of Accelerators for ADS and Comparison with experimental data of Accelerators,” International Topical Meeting on Nuclear Research Applications and Utilization of Accelerators, IAEA, Vienna, Austria, 2009.

[3] R. B. Randall, Vibration based condition monitoring, Wiley, Hoboken, NJ, 2011.

[4] S. Narasimhan, “Automated Diagnosis of Physical Systems,” Proceedings of ICALEPCS07, Knoxville, Tennessee, USA, 2007.

[5] R. Kothamasu et al. System health monitoring and prognostics–a review of current paradigms and practices, Elsevier, 2009.

[6] J. Sikorska et al., “Prognostic modeling options for remaining useful life estimation by industry,” Mechanical Systems and Signal Processing, vol. (25)2, pp. 1803–1836, 2010.

[7] S. L. Gold, “Klystron Gun Arcing and Modulator Protection,” 5th Modulator-Klystron Workshop for Future Linear Colliders (MDK-2001)

[8] M. Pecht and R. Jaai, “A Prognostics and Health Management roadmap for Information and Electronics-rich Systems,” Microelectronics Reliability, 50 (2010) 317–323

Olog, A New Logbook for EPICS!


Olog is a new electronic logbook featuring modern architecture and several client (viewer) options. The recently developed web client is highly customizable and has an adaptive, dynamic Javascript-based user interface allowing it to be used on both desktops and mobile devices.

The Olog module is part of the Distributed Information Services for Control Systems (DISCS) collaboration between BNL, FRIB, Cosylab, IHEP, and ESS. The Olog module is being collaboratively developed by BNL, FRIB and Cosylab.

Olog functions on a client-server basis. The client and server are decoupled and are connected through a REST interface. The server [1] is written in Java and its role is to store and deliver the data and to ensure security. In addition to the web client, a plug-in for Control System Studio (CSS) is also available [3]. It provides good integration with the rest of the CSS ecosystem.

Since the back-end functionality is exposed as a REST service, applications could serve as clients. Currently the REST API client is available as Java and Python library.

The web client [2] is written in Javascript, using the jQuery framework. No other library/framework dependencies are used at the moment. There are three main features/views:

  • browsing existing entries
  • adding new log entries
  • editing an existing log entry


Panes in all three views can be re-sized to suit the user's needs. The home view responds appropriately to screen resolution changes. If Olog is loaded in a browser with a lower screen resolution, the position of the panes will change to accommodate the smaller screen width.


 olog view

Figure 1: Browsing existing log entries

Browsing occurs in the Home view, where details of log entries are displayed according to selected search and filter criteria. Log entries can be filtered according to the logbooks that they were assigned to and according to assigned tags. It is also possible to filter according to time (created in last minute, last hour, last day, last week and created during a specified period). (Figure 1 – left pane)

Results can be restricted further by entering search criteria into the available search bar. Wild-card searches are supported for greater flexibility.

The list of search/filter results consist of a few lines of the description and thumbnails of attached image files (Figure 1 – middle pane). Opening an entry then displays logbooks and tags assigned, description, attachments, properties and other useful data (Figure 1 – right pane).

Each log entry has a specific URL that can be used for sharing log entries between users. The list of log entries is automatically refreshed which is especially useful when more than one user is using Olog at the same time.


 olog create

Figure 2: Adding a log entry

New log entries are easily added, with the only requirement that each log entry must belong to at least one logbook. Logbooks and tags can be assigned by entering their name in an input box with the help of auto-completion or by selecting them from a displayed list of existing logbooks and tags. (Figure 2 – right pane) Olog supports multiple selections allowing more than one logbook and tag to be selected.

After creating the log entry description, a user can also add attachments (text files, images, etc.). Attachments can be added from a file browser, by dragging and dropping them from the desktop or by Olog is a new electronic logbook featuring modern architecture and several client (viewer) options. The recently developed web client is highly customizable and has an adaptive, dynamic Javascript-based user interface allowing it to be used on both desktops and mobile devices.

When modifying log entries all changes are saved in the database but currently only the latest change is shown to the users.

 0000009382-CSS new big

Figure 3: Olog Client in CSS


Only registered Olog users with sufficient permissions are allowed to add or modify log entries. Permissions are managed by groups – users belong to a group that has specific permissions throughout Olog.


Olog is a work-in-progress but promises to become the EPICS logbook of choice!





Cosylab T-shirts are also Popular in China!


We were setting up our conference booth during IPAC 2013 in Shanghai, China at the same time as the students' poster session was going on in the same hall. Therefore many Chinese students "pillaged" our T-shirt supply even before the booth was up. In a flash of intuition, we proposed a deal to all who came asking for T-shirts: "they must wear them the next day and come to our booth for a photo". Not all honoured the deal, but here are two nice ones who did.


The guy next to the woman may not look Chinese to you, and it is indeed Klemen Žagar, the CTO of Cosylab. However, he was so impressed with China that he is seriously considering moving to China to set up a Cosylab branch there. He and his wife (and 2.5 year-old son!) have already started taking Chinese lessons. Since he is wearing a Cosylab T-shirt, we thought he deserves to be in the photo as well.

It is not only Chinese students that like Cosylab T-shirts, Chinese researchers also like them! The (group) photo shows Prof. He Yuan, the leader of the ADS accelerator project at IMP Lanzhou, together with two researchers, Feng Zhu and Shixiang Peng from Beijing, at the Cosylab booth at the LINAC 2012 conference in Tel Aviv. Shixiang Peng was friendly enough to wear the T-shirt the next day and pose for our photographer.


back to previous content

Download printable version of Control Sheet no.16 here.