Beam loss plane recognition for the LHC Gianluca Valentino - PowerPoint PPT Presentation

Beam loss plane recognition for the LHC Gianluca Valentino University of Malta, Msida, Malta Belen Salvachua CERN, Geneva, Switzerland Thanks also to the LHC collimation team for support in measurements and analysis ICFA Mini-Workshop: Machine Learning Applications for Particle Accelerators SLAC, Menlo Park, CA, USA, 1 st March 2018

Outline • Background and motivation: LHC collimation and loss maps • Available datasets • Feature selection • Classification results using NN and GBC • Conclusions Gianluca Valentino Beam loss plane recognition for the LHC 1

Background: LHC collimation • The LHC is equipped with a multi-stage TCL.4R5 TCL.5R5 TCL.5R5 TCL.4L5 TCL.6R5 TCL.5L5 TCL.6L5 IP5 collimation system to protect it from normal and abnormal beam losses. IP4 T T 5 CMS 5 C R IP6 6 C R R T T 4 4 4 A 6 P . P V . A . R H V Q 4 H P B . P D 4 A . . T C 6 T Q 4 L T R C D L 5 C 4 • Normal losses: ensure that proton leakage to TCSP.A4L6 C A T TCDQA.B4L6 5 T T P . S TCDQA.A4L6 C T superconducting magnets is minimal, preventing TCLA.A7L7 TCP.D6L7 TCLA.D6L7 TCP.C6L7 TCLA.C6L7 quenches TCP.B6L7 TCLA.7R3 TCLA.B6L7 TCSG.A6L7 TCLA.6R3 TCLA.A6L7 TCSG.B5L7 TCLA.B5R3 TCSG.6L7 TCLA.A5R3 TCSG.A5L7 • Abnormal losses: protection against fast failure TCSG.E5L7 TCP.6R3 TCSG.D4L7 TCSG.B5R3 TCSG.D5L7 TCSG.5R3 Momentum TCSG.B4L7 TCSG.A5R3 TCSG.B5L7 Betatron TCSG.A4L7 TCSG.4R3 scenarios such as asynchronous beam dump TCSG.A4L7 cleaning IP3 IP7 cleaning TCSG.4L3 TCSG.5L3 TCSG.A4R7 TCSG.A4R7 TCSG.A5L3 TCP.6L3 TCSG.B4R7 TCSG.B5R7 TCSG.B5L3 TCSG.D4R7 TCSG.D5R7 TCLA.A5L3 TCSG.A5R7 TCSG.E5R7 TCLA.B5L3 TCSG.B5R7 TCSG.6R7 TCLA.6L3 TCSG.A6R7 TCLA.A6R7 TCLA.7L3 TCP.B6R7 TCLA.B6R7 • The collimation system cleans particles with TCP.C6R7 TCLA.C6R7 TCP.D6R7 TCLA.D6R7 TCLIB.6R2 TCLIB.6L8 TCLIA.4R2 TCLIA.4L8 TCLA.A7R7 large betatron and off-momentum offsets ALICE LHC-b TCTPH.4L8 TCTPH.4R2 TCTPV.4L8 TCTPV.4R2 1 T 1 T R C C R 4 T 4 T IP2 . IP8 H . P V P P H P V TCTPV.4L2 T . . T 4 ATLAS 4 C C L L T TDI.4L2 T 1 C TCTPH.4L2 1 T T T D P T I . V C 4 . T R 4 P R 8 H 8 . 4 TCL.6R1 TCL.5R1 R TCL.6L1 TCL.5L1 TCL.4R1 TCL.5L1 TCL.4L1 8 B1 B2 IP1 Gianluca Valentino Beam loss plane recognition for the LHC 2

Background: LHC collimation A double-sided LHC collimator (can Multi-stage halo cleaning process be installed to clean in the horizontal, vertical or skew planes) Absorber (TCLA) Primary collimator (TCP) Tertiary collimator Secondary (TCTP) collimator (TCSG) Gianluca Valentino Beam loss plane recognition for the LHC 3

Background: beam losses • In order to monitor beam losses, ~3600 Beam Loss Monitoring (BLM) ionization chambers are placed around the LHC. Ionization chambers Gianluca Valentino Beam loss plane recognition for the LHC 4

Background: beam loss maps • The betatron cleaning system is qualified by intentionally creating high losses in the horizontal (H) or vertical (V) planes using the transverse damper. • Cold = losses in superconducting magnets (arcs) • Warm = losses in normal magnets / IRs • Collimator = losses at BLM at collimator IR7 IR3 • Loss maps are generated during test fills with low intensity (few bunches). • With the collimation system properly set up, we expect highest losses at the bottleneck in IR7. ~3600 BLM readings around the ring Gianluca Valentino Beam loss plane recognition for the LHC 5

Background: beam loss maps • We can blow-up individual bunches in a given beam & plane: • A zoom into IR7 gives a better view of the hierarchy: Gianluca Valentino Beam loss plane recognition for the LHC 6

Problem definition & motivation • The objective is to be able to automatically classify between the four types of loss planes: • Beam 1 Horizontal (B1H) • Beam 1 Vertical (B1V) • Beam 2 Horizontal (B2H) • Beam 2 Vertical (B2V) • Therefore our problem has 4 output classes. • Understanding the beam loss characteristics and dynamics during normal operation is crucial to correct them and understand their long-term impacts e.g. R2E effects Gianluca Valentino Beam loss plane recognition for the LHC 7

Problem definition & motivation B1H B1V B2H B2V Gianluca Valentino Beam loss plane recognition for the LHC 8

Available datasets • The data from the ~3600 LHC BLMs were extracted each time the transverse damper blow-up was running during 2017. • This resulted in 5893 loss maps, which were then narrowed down based on: • Beam intensity loss: the intensity loss in each loss map should be > 1E8 protons to have sufficient resolution; • Collimator settings: the collimator positions have to be identical in each loss map; • Visual checks: to ensure a correct hierarchy was in place. Gianluca Valentino Beam loss plane recognition for the LHC 9

Available datasets • As the beam parameters, collimator settings etc are different between injection and flat top, separate models were trained for both cases: Beam & Plane Injection Top Energy B1H 84 496 B1V 132 599 B2H 127 383 B2V 123 129 Gianluca Valentino Beam loss plane recognition for the LHC 10

Feature selection • Two feature sets were considered: 1. All the BLMs in the IR7 longitudinal position range (19400 – 20600 m): 261 BLMs. 2. Only the BLMs located at collimators in the same longitudinal position range: 41 BLMs. • The BLM readings in each loss map were normalized to the same BLM (TCP.A6L7.B1 @ 19796 m), which generally gives the highest readings for B1 loss maps. Gianluca Valentino Beam loss plane recognition for the LHC 11

Training Procedure • Two separate models were trained for the loss maps: • at injection energy (450 GeV) • at top energy (6.5 TeV) • Loss maps in each of the four datasets (B1H, B1V, B2H, B2V) were split using a 50:50 ratio between training and testing datasets. • The allocation of a particular loss map to the training or testing datasets was done randomly. • The models were then use to predict labels for the as yet unseen testing dataset, which were compared to the original ground truth. • The final classification success rate was calculated by averaging the prediction performance on the testing dataset over 5 tries (test + train) to avoid a lucky split. Gianluca Valentino Beam loss plane recognition for the LHC 12

Results with a Neural Network 4 hidden layers: (300, 130, 75, 30) Beam & Plane Injection Top Energy B1H 98.1% 99.8% Using only the B1V 99.7% 99.9% collimator BLMs B2H 96.9% 98.5% B2V 97.7% 96.9% Beam & Plane Injection Top Energy B1H 76.6% 99.4% Using all IR7 BLMs B1V 91.8% 98.5% B2H 96.8% 99.5% B2V 96.5% 96.6% Gianluca Valentino Beam loss plane recognition for the LHC 13

Results with Gradient Boosting Classifier n_estimators = 1000, max_depth = 20 Beam & Plane Injection Top Energy B1H 97.6% 99.0% Using only the B1V 98.2% 99.5% collimator BLMs B2H 95.3% 99.5% B2V 97.1% 96.3% Beam & Plane Injection Top Energy B1H 97.6% 100% Using all IR7 BLMs B1V 99.1% 99.2% B2H 95.9% 99.1% B2V 96.5% 96.2% Gianluca Valentino Beam loss plane recognition for the LHC 14

Classification of losses during LHC operation • Loss maps are performed in controlled conditions (beam and plane is known), and ML models were trained on these data. • Next step: applying the models trained on loss map data to actual losses in operation. • Two scenarios considered: • Long-Range Beam-Beam (LRBB) beam test • Losses during the standard LHC machine cycle Gianluca Valentino Beam loss plane recognition for the LHC 15

Existing beam loss decomposition method • Based on SVD (see M. Wyszynski, CERN summer student project & B. Salvachua et al., “Decomposition of beam losses at LHC”, IPAC’17). • Works for both off-momentum and betatron losses. • Uses a calibration factor (obtained through dedicated collimator scraping measurements) to convert BLM readings in Gy/s to proton/s. • A subset of only 6 BLMs per beam, at H and V collimators, is used. This vector is then decomposed as a linear combination of the individual B1H/B1V/B2H/B2V contributions. • It is static and not easily adaptable to new machine configurations (requires manual selection of BLMs). Gianluca Valentino Beam loss plane recognition for the LHC 16

Long-Range Beam-Beam test • Beam test using wires installed in collimators to compensate the octupolar term of the beam- beam in IR5. • There was an initial B2H blow-up, followed by B2H B2H additional losses in B2H as the wires were B1H B2H switched on and off. • The ML algorithm correctly classified the 3 spikes in losses which were ~1e8 p. There was one misclassification (though here the losses in B1 and B2 are ~equal & ~1e7 p). *classification result Gianluca Valentino Beam loss plane recognition for the LHC 17

Losses during the squeeze in standard operation GBC SVD Future work: change to more “hierarchical” classification (i.e. consider that we can have losses simultaneously in B1 and B2) Gianluca Valentino Beam loss plane recognition for the LHC 18