List of questions

Can we produce thicker sheets or bulk material of grain oriented steel, and control the grain orientation?

Grain-oriented electrical (GO) steel is produced to obtain magnetic properties superior to normal steel. Currently it is produced in the form of sheets with a linear orientation of the grains and with a thickness of less than 2mm. The sheets are packed into laminates, cut and assembled together in applications requiring an optimal performance ratio between the magnetic field and the electrical current, such as transformers and electrical motors. To avoid eddy currents, the sheets are generally coated. GO steel is also attractive in special high-field electromagnets in difficult environments where the space for coils is very limited or cooling is not desirable. However, the limited thickness and the linear grain-orientation severely constraints the use and challenges large applications. In addition, the magnetic properties of GO steel is highly sensitive to excessive bending, heating as in welding, and mechanical stresses. The SHiP Collaboration at CERN is currently working on the development of a large magnetic particle sweeper based on six five-meter long magnets with a total magnetic core mass of >1500 tonnes. The transverse dimensions of the magnets are of the order of several meters. The application requires highest possible field gradients (>1.6T), very narrow gaps between the core field and the return field, and an overall complex field shape. For this reason, the application leaves very limited room for coils, no room for cooling, and the support structure should be minimised. The packing factor is essential wherefore gluing sheets is not an acceptable solution. Currently the R&D is investigating the use of 300um sheets which are packed and welded together. Annealing at high temperature to remove the formation of carbon structures is studied to restore the magnetic properties in the welded regions, requiring very large ovens.

How to produce, cut and polish radiation-hard garnet crystals more efficiently for large detector applications?

Future detectors built to explore the energy frontier and the intensity frontier of physics will be operating at very high beam collision rates. As a consequence the detectors will need to provide significantly better precision timing information and granularity than today's detectors. In addition they will need to be tolerant to very harsh radiation environments. For these reason, there is an active R&D ongoing in radiation hard, scintillating crystals coupled to compact semiconductor photodetectors. In particular aluminium garnet crystals co-doped in different ways to increase light yield and shorten the response time, as well as improving the radiation hardness, show very promising results. The construction of complex detectors, such as calorimeters, requires producing thin and long crystals with very high demands on purity, uniformity and polishing to guarantee light transmission and signal speed. These requirements and the large quantities make the application of crystals financially challenging. Breakthrough in the production and processing of garnet crystals would bring very large benefits to detectors both in fundamental research and in industrial applications.

How to construct efficiently large and complex detector absorbers from tungsten alloys, whose composition are driven by the physics application?

Particle energy measurements is based on the technique of absorbing the particle by provoking it to interact in a dense medium. In this process the energy released is captured by integrated active detector components that sample ionisation or scintillating light from the shower of secondary particles. The energy measurements in future detector is severely challenged by the ambitions to further increase beam collision rates. The effect of high beam collision rates is a huge pileup of interactions in the calorimeter absorber that must be disentangled at the level of their location and timing. Overlap between the secondary particle showers can be reduced by increasing the density of the medium and by a fine sampling of the showers allowing reconstruction of the shapes and extent of each shower. Along these lines several projects are underway that investigate the use of absorbers made from tungsten, or tungsten alloys whose composition is optimized with the physics performance, and active sampling elements consisting of scintillating crystals. The structure is relatively complex, requiring new techniques for the construction of the absorber in order to integrate the active elements.

Radiation hardness on greases: Is there a roller screw/lubricant (dry) system that can withstand the conditions in a radiation environment, and take up to 10MGy?

Several equipment (collimators, beam dumps, beam targets, etc.) installed in the accelerator are equipped with roller screws / jacks / actuators. The accelerator environment is made of radiation (High energetic mix-field particles) and a very smooth air circulation guarantying constant temperature (21C) and hygrometry (~70%). The maintenance of such equipment is not easily possible, maximum once a year and without dismounting such equipment from the machine. Such equipment shall be 100% reliable as they have a direct impact on the accelerator operation.
For instance, for collimators, we use roller screws and we regularly need to re-grease it (1 every 2 years) as we observed (visually) on some of them a dry aspect of the screw. This situation is not really acceptable because greasing collimators cannot be done by robots and operators a taking dose when each time they perform this maintenance. We are investigating, for years now, different roller screw design (re-circulation, planetary roller screws, etc) that would work with dry lubricant (BALINIT® WC/C (first “hard” layer) + DICRONITE® DL-5 (second “soft” layer)) hoping such lubricant would be more radiation hard (up to 10 MGy) and would guaranty a good behaviour of the screw without any maintenance. We are progressing but we don’t have a final solution yet.

Is there a method to heavily bend 316L tubes (6mm or 18mm) with nearly no deformation?

In HIP designs we do at CERN, we propose to insert 316L SS tubes (6mm or 18 mm diameters) in grooves, precisely machined in Cu ally blocks. Tubes are bent, (180 degrees) and during this process they get deformed, leading to some difficulties in the tube fitting into the grooves. As a direct consequence, there are gaps (1 to 2 mm for 18 mm diameter tubes) between the tubes and the groove that generates problems during the HIP. To solve that problem, CERN is precisely measuring the tubes’shape after bending, and is machining the groove according to the exact tube shape. This process could be prevented if the bending of the tube would generate nearly no deformation.

Can we design a cooling solution in a vacuum chamber that does not include welded seams?

Each time we need to design a new equipment which has to be installed in the accelerator vacuum (Vacuum levels between 10-9 mbar and 10-11 mbar) and which need efficient cooling, we are facing a serious design constraint: the cooling ducts should be without any welded seams as it is a potential source of leak in the vacuum. We would be very interested in finding designs which guarantee the leak tightness while allowing to efficiently cool down the parts located into the vacuum chambers (We deal with equipment extracting few kW to several hundreds of kW of power). So far the HIP technology is very interesting but presents some constraints in the manufacturing/implementation. Maybe integrated design using additive manufacturing techniques of several different materials (Stainless steel/Copper or other material associations with similar or very low CTEs) may enlarge the possibilities?

How can we make industrial robots lighter, while maintaining their precision and dynamics?

Increasing only the payload/robot-weight ratio is a ask often very well addressed by
cable-driven robots. Though, since robots should fulfill certain tasks, a lot of other requirements
and constraints are arising, which are all somehow coupled (often indirect proportional) to the
payload/robot-weight ratio. The challenge is to optimize all or more of those parameters at the
same time.
Parameters could be:
1. Payload / Robot-Weight: Good (High) values for cable-driven robots and very poor (low)
values for snake like robots.
2. Workspace / Robot-Weight: This ratio is used to classify robots in normal and lightweight
(above a certain threshold) robots. Good (High) values for mobile robots (→− ∞) and poor
(low) values for industrial pallet robots (high payload in small workspace).
3. Workspace / Robot-Space: This factor can be used to describe how much fixed support
structures a robot uses. Robot-Space is considered the space which is necessarily occupied
by the robot to reach a certain position and orientation of the endeffector. This yields good
(high) values for snake like robots (→− ∞) and poor (low) values for cable-driven (∼ 1) robots.
4. Dexterity: Highly redundant mechanical structures like snake robots will provide outstanding dexterity, but poor performance in terms of payload.
5. Accuracy: High positioning accuracy and repeatability will be achieved by rigid robots with
low vibrations, but lead to a bulky and heavy design.
6. Dynamics: High dynamics will limit workspace, payload and require powerful (heavy) motors.

How can we increase safety for humans in close human/robot collaborations?

The main challenge of human/robot collaboration lays in the basic principle of robotic arm and mobile robot design. The main performance measure of today’s robots are payload capacity, speed, robustness and precision. To achieve high payload capacity and speed robots must be outfitted with high-torque actuators. For example a standard industrial robot such as Kuka KR10 has a combined
payload and mass of 64 kg. The robot motors must be able to accelerate this enormous
loads.
In order to make the robotic arms robust, most industrial grade robot arms are made of metal.
The outside shielding and the robot links are both metal, even tough the shielding could be
easily made from lighter materials, such as plastic of carbon-fiber composites. The
combination of high mass and high velocities results in a very high kinetic energy while
moving. Additionally the metal shielding is very stiff, which means that when it comes in
contact with the environment it transfers its kinetic energy rapidly and efficiently. This results
in huge damage to whatever it comes into contact with. Just imagine hitting a ball with a
baseball bat or mittens, even if the kinetic energy is the same, the mittens are less
dangerous.
To realize good precision in robotic movements the feedback controllers of the robot must
have high gains. This will result rapid movement with high disturbance rejection. This
disturbance in structured, ideal, come from the weight of the object manipulated.
However when the robot comes into contact with a person it will also recognize it as a
negative outside disturbance, and it will counteract it with higher feedback. This means
that when a robot hits an operator with great speed and impulse it will not slow down, on
the contrary, it will increase its force. This is the reason why industrial robotic arms are so
dangerous to work around.
For robots to be able to coexist with humans several design changes must be carried out.
Softer materials must be used for their coating to reduce the impact of collisions. Their
mass must be reduced to lower their kinetic energy (this is also very useful, because less
power is needed to move the arm, and less expensive motors can be used). Most
importantly the control of the robotic arms must be changed. They have to be able sense
the forces introduced by the environment, to be able to slow down when coming in contact
with a person. Additionally, vision systems with artificial intelligence must predict and plan
to avoid collision with the operators and workers.

What can Sweden do for helping CERN to develop a canted-cos-theta dipole magnet for the LHC?

Due to the High Luminosity upgrade of the Large Hadron Collider (HL-LHC), the orbit corrector dipoles (MCBC and MCBY) located just before the interaction points of the accelerator will receive significantly increased gamma radiation doses, i.e. up to 20 MGy over the HL-LHC lifetime requiring to change up to 32 out of the 122 installed magnets of this type in LHC. Since these magnets were not designed to withstand such high gamma doses, new magnets have to be designed and produced using radiation hard impregnation insulation and materials. One solution is to procure magnets based on a very similar design as currently used in the LHC using more radiation hard materials. An alternative, likely to be much more radiation hard and cost efficient, is the manufacture of Canted Cosine-Theta (CCT) magnets with similar operating parameters. However, in this design, a number of outstanding, not yet solved, challenges exist; mainly the magnet protection and reaching a similar range in the operating current. For a proof of principle a fully-fledged design would need to be elaborated and a full scale prototype be built. In this challenge, we want to discuss how Sweden and Swedish Industry could do the design, manufacture and test of a full-scale twin-aperture canted-cos-theta prototype.

How can we develop Superconducting Magnet Energy Storage (SMES) for the LHC at CERN?

It takes of the order of an hour to fill the Large Hadron Collider (LHC) at CERN with the two colliding proton beams, accelerate them to full energy and focus them to collide. The beams will then circulate in collision mode for 10-20 hours providing proton-proton collisions to the experiments at the collision points. A rapid variation, a ‘glitch’, in the electrical power supplied to the LHC of a duration of only a fraction of a second is enough to lose the beams. The long time needed to start up the LHC again implies significant loss of the valuable beam-collision time. To prevent this to happen there needs to be a very large electric-energy storage available that can very rapidly, within a fraction of a second, compensate for the glitch by delivering a very high current over a period a second. This can be realized using the Superconducting Magnet Energy Storage (SMES) technique. CERN has expressed interest in developing in collaboration with a Member Country a large-current superconducting-magnet system, with stored energy in the range of 10 Mega-Joules, and capable to deliver a large fraction over a time scale of half a second. About 10 such systems would be needed for the LHC. Each magnet will require a dedicated power converter. The integration of the magnets and their power converters in the LHC complex will require extensive system engineering. The FREIA Laboratory in Uppsala, which develops and tests cryogenic equipment for the LHC as well as for the European Spallation Source (ESS), is looking for a Swedish industrial partner to design, fabricate and test a SMES prototype system for CERN. Such a prototype would be of great interest for the potential use of SMES systems also at the large synchrotron light sources (like MAX IV), tokamaks (like ITER) and future colliders (like FCC).

How can we fabricate -53 degrees CO2 cooling systems for the experimental setup at the ATLAS experiment?

The ATLAS experiment at the CERN Large Hadron Collider is developing a new so-called Inner Tracker which, based on semiconductor technique, will measure the tracks of all the charged particles being produced in the collisions of the two proton beams and this in a location very near the vacuum tube inside which the two proton beams collide. The Inner Tracker needs to be cooled to temperatures down to -40 degrees in order for its sensors to survive the HL-LHC radiation levels. The ATLAS experiment is located in the LHC tunnel at ca 100 m below ground. The plan is to have a primary refrigeration circuit located at the surface and bring the CO2 used as refrigerating medium down to the level of the experiment (100 mt below ground) using high pressure pipes. There, a secondary CO2 cooling circuit will be cooled by the primary CO2 circuit using heat exchangers. It has been estimated that, in order to safely keep ca -40 degrees of evaporation temperature in the experiment with the secondary circuit underground, the temperature of the CO2 in the primary refrigeration circuit has to evaporate CO2 around -53 degrees. Towards the end of the current year CERN will issue a tender for about 10 such primary refrigeration circuits for which the current estimated order value is about 4.5 MCHF. The FREIA Laboratory in Uppsala, which is currently developing and testing cryogenic equipment for the LHC as well as for the European Spallation Source (ESS), is looking for a Swedish industrial partner to design, fabricate and test primary CO2 cooling circuits of the required kind for the ATLAS experiment.

2745: How can we increase the “human touch” for robots working with humans in Big Science?

2748: How can we increase proprioception in maintenance teleoperation in big science facilities?

2745: Interventions requiring objects manipulation are performed by commanding industrial-commercial grippers, used as end-effector on robotic arms, in velocity, without having knowledge of the applied force. When the objects to be handled or manipulated are made of high rigid materials, such as steel, they do not need a fine control to regulate the applied force.
In case of fragile objects these interventions will be more difficult to implement without a control that manages the force applied by the grippers. Thus, the object could be broken and human intervention in hazardous environments would be necessary. An idea might be to equip the grippers with i) force sensors in order to obtain information about touch and slippage between fingers tip and the grasped object surface, and ii) an automatic control could solve the problems described above. Furthermore, force, touch and slippage information could be used to return a haptic feedback (e.g. vibrotactile feedback) to operator during the teleoperated task.


2748: Large facilities, which have hazard of ionizing radiation produced during operation, require teleoperation of maintenance tasks. This is due to for example the interaction of the generated beams with matter and its activation and therefore its inaccessibility to humans.
Therefore, robot deployment in such big science facilities requires reliable sensor feedback and a certain independency of the incorporated system. To ensure full availability, state of the equipment must be sensed thoroughly. The robot’s battery status is the most obvious example, where the system would seek for a charger. A deployed robot that is equipped with feedback to sense eventual damages on the machine, can prevent itself from being affected using for example electric proprioceptive sensors. If the system is deployed statically, heat or radiation monitoring can be of necessary to ensure operation within the activated areas and can account to faults in the equipment due to accumulated dose effect damages. With increased proprioception, preventive maintenance can be addressed to ensure the effective operation of the facilities.
The deployed robotic systems typically must adapt to new tasks and environments for example by changing its morphologies i.e. robot configurations. Current state of the art technologies for example show leveraged machine learning algorithms to derive the precise bending and twisting of soft robotics limbs from the analysis of diffuse reflected light through embedded optical fibres. [1]
Another example can be virtual fixtures could help the robot to avoid collisions if visual perception is integrated. Further, proprioception can be established by using machine learning for autonomous operations and deep learning for object and pose recognition. This could be used in such an environment, since the structure can be explored and mapped beforehand, and the machine could learn from every task performed.
Due to the diversity of the tasks, the lack of robot friendly structures within the facilities makes operation sometimes become a challenge. Like most of the times, the usage of robots is not integrated at the facilities design phase. Thus designing machines that can be maintained by robots using appropriate and easily accessible interfaces will increase the availability and decrease human exposure to hazards.

[1] https://www.eenewseurope.com/news/soft-robotics-proprioception-let-machine-sort-it-out

2724: How can we make use of drones more efficient and more compatible in terms of flying time and having them work autonomously?

2769: How can we use drones for monitoring in the accelerator tunnels and other hostile environments?

2724: Current drone applications are limited mainly by autonomy and accessibility of narrow zones that need aggressive flight techniques. In Big Science Facilities, the areas to inspect and to maintain are wide and they need drones that are able to fly autonomously over long distances. In addition, in areas with the risk of contamination, drones propellers should generate very low wind turbulences to avoid contaminated dust being moved in different facilities regions.

2769: Many facilities are now being build with no windows and cameras only and drones could be used to facilitate new camera angles. We have to consider the radioactive environment in these environments and specifically the radioactive dust that regular drones stir up. Can vi reduce dust disturbance? What are the specifications of new drones that are needed to fit these environments?

How do we optimize the flow of data in machine learning projects? 

In all machine learning projects, it is critical to have access to large amounts of well-organized data to feed the learning process. Every Big Science facility´s control systems will generate at least an order of magnitude larger volumes of control system and automation related data than typical existing, large industries. This rich mix of data will range from typical industrial process control systems to state-of-the-art control and data acquisition systems. So How do we optimize the flow of data? Who are to choose the data that is shown? How do we assess the quality of the data? How much data do we need for processing and visualization? What should be sorted out? Where should the data be processed? Where and how do should we store it? Should all data be stored, how to sort data?

How do we develop Intelligent Alarm Handling ?

Alarms are important for process-understanding, availability, reliability and maintainability. The chain reaction of a triggered alarm are not always the original source of the malfunction in the system. How do we optimize advanced alarm suppression techniques to meet the required availability and reliability goals.

How to create a Software Development Ecosystem for Machine Learning?

I a few years from now many developers and system experts throughout the whole organization as well as in the collaborating institutes will use machine learning models to operate the ESS machine. This will imply that within a few years we can expect to have many ML models deployed in production. To facilitate an ecosystem maintaining a large set of ML models in a highly complex facility will require structure, processes and tools. This raises several questions:

How do we assure successful operation cycles (start-up, machine operation, maintenance, shut-down and machine upgrades)? More specifically, how do we keep track of deployed ML models and how do we know which ML models that need to be re-train, re-test and re-validate after maintenance, machine upgrades or other changes?

How do we assure robustness of machine learning models? How do we minimize the need of updating ml-models after e.g. maintenance?

How do we combine physics models and machine learning models?

What processes and tools do we need to assure quality of ml models before these are allowed to be deployed in production?

How can we optimize machine performance with machine learning?

We rely heavily on physics models and simulations to determine the optimum parameter settings of the machine, especially during start-up. However, with an expected number of 1.6 millions parameters to adjust and the fact that reality is always more complex than any theoretical model can predict, physicists, engineers and machine operators spend lot of time optimizing the machine performance by trial and error.

What does tomorrows control rooms look like? 

How does human and machines cooperate? How can we use the gaming industry for visualization in new ways in the control room. How do we incorporate robotics in complicated environments. How do we integrate it with VR (Virtual reality) and where should we use (AR) Augmented reality?

How do we together drive the development of future Control Systems for Complex Processes (EPICS / Tango)?

The research facilities are control using control system software like EPICS and Tango. The development of these platforms as these are essential for the future operations. How can we drive the development for our focuses in these platforms and can we get the industry to get involved and where do we share common interests.

2724: How can we make use of drones more efficient and more compatible in terms of flying time and having them work autonomously?

2769: How can we use drones for monitoring in the accelerator tunnels and other hostile environments?

2724: Current drone applications are limited mainly by autonomy and accessibility of narrow zones that need aggressive flight techniques. In Big Science Facilities, the areas to inspect and to maintain are wide and they need drones that are able to fly autonomously over long distances. In addition, in areas with the risk of contamination, drones propellers should generate very low wind turbulences to avoid contaminated dust being moved in different facilities regions.

2769: Many facilities are now being build with no windows and cameras only and drones could be used to facilitate new camera angles. We have to consider the radioactive environment in these environments and specifically the radioactive dust that regular drones stir up. Can vi reduce dust disturbance? What are the specifications of new drones that are needed to fit these environments?