1 Investigation of Geometry Modeling Methodfor ATLAS Simulation Niko Tsutskiridze Georgian Technical University European Organization for Nuclear Research CERN SCCTW’2016, 06/10/2016

2 LHC Machine at CERN ATLAS Detector length ~40 m, height ~22 m,weight ~7’000 tones ALICE Detector LHCB Detector CMS Detector

3 Research Hypothesis Several reasons can cause discrepancies between Data and Monte-Carlo. Several investigations show that they are coming by the reason of geometry descriptions in simulation It is possible to predict 2 hypothesis why faults are exist in geometry descriptions: Hypothesis #01: Inaccuracies added by geometry transactions within the simulation software infrastructure Hypothesis #02: Discrepancies between the design and the geometry implementation inside the simulation

4 Geant4 Toolkit GEANT4 is a platform for simulation of facilities and physical events by modelling of the passage of particles through the matter GEANT4 implementing in High Energy, nuclear and Accelerator physics as well for studies in medical and in space science G4DNA G4MED G4EMU G4NAMU BABAR BOREXINO LHC GEANT 4

5 Geometry Simulation LoopSeveral Chains have been developed: GEANT-to-CATIA GeoMODEL-to-CATIA CATIA-to-XML CATIA-to-GeoMODEL

6 Investigation of Simulation InfrastructureChecking Hypothesis 01: Investigation of Simulation Infrastructure

7 Objectives of AnalysesXML GeoMODEL GEANT-4 Interpretation Engine T1 T2 Categorization of geometry of Detector components Selection Methods for description Test runs of test examples Case study of transactions Systematization and learning of results

8 I. Categorization of Geometry

9 I. Categorization of GeometryGeometric Primitives Shapes without cuts Both regular/irregular shapes Both convex/concave shapes Combined Objects Grouping components with cuts Typical Joining Grouping components with typical joining’s 22 typical primitives have been separated 33 combined objects have been separated 29 combined objects with typical joining’s have been separated

10 I. Categorization of Geometry1st class Geometric Primitives 22 Total: 84 2nd class Combined Object 33 3rd class Typical Joining 29 Conclusion: ATLAS detector geometry can be described by 84 typical representors of class of objects

11 II. Selection of Methods for Description

12 II. Selection of Methods for DescriptionBox Cone Conical Section Cylindrical Section or Tube Parallelepiped  Trapezoid  Generic Trapezoid Sphere, or a Spherical Shell Section Solid Sphere Torus Polycons Polyhedra  Tube With an Elliptical Cross Section Ellipsoid  Cone With an Tube With a Hyperbolic Profile Tetrahedra   Box Twisted Tube Section AGDD/XML Cube Tube Pyramid Cylinder chain Arbitrary Symmetric Double Symmetric Geant4 GeoModel Box Cone Parallelepiped Polycone Polygon Trapezoid (Complex) Tube Tube Section (Simple) 12

13 II. Selection of Methods for Description

14 II. Selection of Methods for DescriptionFinally, for all above selected typical representatives of object classes of ATLAS detector, full set of possible methods of description were selected: 1st class of 22 objects: 4’460 methods 2nd class of 33 objects: 6’579 methods 3rd class of 29 objects: 4’636 methods Total: 15’675 methods

15 II. Selection of Methods for DescriptionCriteria #01: Arbitrary_polygon method should be separated as a standalone method, while Geometry description requires minimal number of Boolean operations and Move/Rotation transactions Geometry can be described directly in position by only Z axis displacement and Z axis rotation. I. II. III. Example: Descriptions of Octadecagonal Prism Conclusion: Exclude Methods II and III

16 II. Selection of Methods for DescriptionCriteria #02: Minimization of number of used methods in description Ensure compactness of code Reduce number received clashes, contacts and inaccuracies of positioning Ensure better performance by reducing number of regions for consideration during the tracking Example: Descriptions of Cube with Cut II. I. Conclusion: Exclude Method II

17 II. Selection of Methods for DescriptionCriteria #03: Exclude descriptions which are using same transactions and methods Example: Descriptions of Dodecagonal Prism with Cuts II. I. Conclusion: Either I or II should be excluded

18 II. Selection of Methods for DescriptionCriteria #04: Exclude descriptions with same consequence of methods Example: Descriptions of Icositetrahedronal prism with cuts I. II. Conclusion: Either I or II should be excluded

19 II. Selection of Methods for DescriptionTotal number of methods has been analysed and just unique cases of descriptions were selected: Before Separation After Separation 1st class of 22 objects: 4’460 methods 2nd class of 33 objects: ’579 methods 3rd class of 29 objects: 4’636 methods Total: 15’675 methods 1st class of 4’460 methods: 11 methods 2nd class of 6’579 methods: 38 methods 3rd class of 4’636 methods: 28 methods Total: 77 methods Conclusion: 77 unique examples have been formed for the investigation of quality of geometry transformations doing by simulation software.

20 III. Test Runs

21 III. Test Runs Simulation Loop 77 Test Examples 51 cases with faultsTest Example N Max. inaccuracies 77 Test Examples Simulation Loop 51 cases with faults 26 cases without faults

22 Example of Test Run T1 T2 T3 T4 T5 T6 T7

23 Example of Test Run 1 2 3 4 5 6 7 GeoM ∆1 G-4 ∆2 1 x y z 2 -0.01 -0.02y z 2 -0.01 -0.02 3 0.01 4 5 6 7 Volume 1 2 3 4 5 6 7

24 IV. Case Study of Transactions

25 Sub-Case №01: T1/T2/T4 transactions together with Boolean SubtractionIV. Case Study of Transactions Sub-Case №01: T1/T2/T4 transactions together with Boolean Subtraction Results: GeoM ∆1 G-4 ∆2 1 x y z 2 -0.01 3 -0.02 4 5 0.01 6 7 Volume T1 T2 T3 T4 T5 T6 T7

26 IV. Case Study of TransactionsSub-Case #02: T6 movement together with T1/T2/T4 transactions and T3/T5 Boolean Subtraction Results: GeoM ∆1 G-4 ∆2 1 x y z 2 -0.01 3 -0.02 0.01 4 5 6 7 Volume T1 T2 T3 T4 T5 T6 T7

27 IV. Case Study of TransactionsSub-Case #03: T7 rotation transaction and T1/T2/T4 transactions together with T3/T5 Boolean Subtraction Results: GeoM ∆1 G-4 ∆2 1 x y z 2 -0.01 -0.02 3 4 5 6 7 Volume T1 T2 T3 T4 T5 T6 T7

28 V. Systematization and Learning of Results

29 V. Systematization and Learning of ResultsUsed Methods Transactions Inaccuracies GeoModel Geant4

30 V. Systematization and Learning of ResultsUsed Methods Transactions Inaccuracies GeoModel Geant4

31 V. Systematization and Learning of ResultsUsed Methods Transactions Inaccuracies GeoModel Geant4

32 V. Systematization and Learning of ResultsUsed Methods Transactions Inaccuracies GeoModel Geant4

33 V. Systematization and Learning of ResultsConclusion №01 For all type of detector geometries dimensional, form and positioning faults are caused by Boolean operations 51 Examples with faults 26 Examples without faults 77 Test Examples With Booleans Without Booleans

34 V. Systematization and Learning of ResultsConclusion №02 All internal surfaces received by Boolean subtraction of parametrical primitives from Box brings 0 faults Test Example #09 Test Example #15

35 V. Systematization and Learning of ResultsConclusion №03 Boolean operations are correlate with Move and Rotate transactions executing after the Boolean. All Move/Rotate transactions before Boolean are fine

36 V. Systematization and Learning of ResultsConclusion №04 For all external surfaces created by subtraction of parametrical primitives from Box, Boolean operation don’t correlated with Move/Rotation transactions Test Example #08 Test Example #56

37 V. Systematization and Learning of ResultsConclusion №05 For some internal surfaces created by subtraction of parametrical primitives from Polygon methods, Boolean operation don’t correlated with Move transactions Test Example #19, #20 Test Example #22 Test Example #38, #39 Test Example #34, #35

38 Conclusions of Hypothesis #11. Hypothesis #01 has been confirmed: The simulation software infrastructure introduces geometrical inaccuracies 2. For all type of detector geometries the faults in dimension, form and positioning are caused by Boolean operations 3. All internal surfaces received by Boolean subtraction of parametrical primitives from a Box result in zero faults 4. Boolean operation inaccuracies are correlated with Moving/Rotation transactions in GEANT4 5. For all external surfaces created by the subtraction of parametrical primitives from a Box, Boolean operation Inaccuracies do not correlate with Moving/Rotation transactions 6. For some internal surfaces created by the subtraction of a Polygon methods via Tube method, Boolean operation do not correlate with Moving transactions

39 Checking Hypothesis 02: Investigation of discrepancies between the design and the geometry implementation inside the simulation

40 Objectives of AnalysesReproduction of Geometrical Model of COIL in CATIA Decomposition and Mass analysis of COIL Compare analysis between CATIA and Geant4 COILs Integration conflict checking

41 I. Reproduction of Geometrical Model of COIL in CATIA

42 Reproduction of Geometrical Model of COIL in CATIASource geometry has been taken from Smarteam Engineering Database 225 manufacturing drawings have been founded on CDD and missing parts was added to primary Smarteam geometry SmarTeam Model CATIA Model A - A A

43 II. Decomposition and Mass analysis of COIL

44 Volume 1. Cryostat Long (Top)Decomposition and Mass analysis of COIL Volume 1. Cryostat Long (Top) Volume 2, 4, 6, 8. Cryostat Corner Volume 3, 7. Cryostat Short Volume 5 Cryostat Long (bottom)

45 Decomposition and Mass analysis of COIL Volume 12. Thermal ShieldingVolume 9. Voussoirs Volume 10. STEFFENERS Volume 11. Ribs Volume 12. Thermal Shielding

46 Decomposition and Mass analysis of COIL Volume 15. Coil casing partVolume 13. Tie Rod Volume 15. Coil casing part

47 Decomposition and Mass analysis of COILVolume 16 Volume 17. Services Volume 18. Supports of Services

48 Volume 19. Supports of CoilDecomposition and Mass analysis of COIL Volume 19. Supports of Coil

49 Volume 20. Ribs of Thermal ShieldingDecomposition and Mass analysis of COIL Volume 20. Ribs of Thermal Shielding Volume 21. Ribs of Coil casing

50 Decomposition and Mass analysis of COIL91’914 kg = 10’088 kg + 1’344 kg kg + 11’368 kg + 12’344 kg + 5’336 kg + 4’824 kg + 2’020 kg + 2’928 kg + 18’ kg kg + 11’ kg kg kg kg kg + 1’873 kg Total mass of COIL- 91’914 kg

51 III. Compare analysis between CATIA and Geant4 COILs

52 Compare analysis between CATIA and Geant4 COILsVolume 1 Volume 2 Volume 3 Volume 5 Volume 9 Volume 4 Volume 11 Volume 8 Volume 7 Volume 6 Volume 12 Volume 10

53 Volume 1. Cryostat Long (Top)Compare analysis between CATIA and Geant4 COILs Volume 1. Cryostat Long (Top) Volume 2, 4, 6, 8. Cryostat Corner CATIA CATIA G4 G4

54 Compare analysis between CATIA and Geant4 COILsVolume 3, 7. Cryostat Short Volume 5 Cryostat Long (bottom) CATIA CATIA G4 G4

55 Compare analysis between CATIA and Geant4 COILsVolume 9. Voussoirs Volume 10. STEFFENERS CATIA CATIA G4 G4

56 Compare analysis between CATIA and Geant4 COILsInner Parts Volume 11. Ribs CATIA G4 CATIA G4

57 Compare analysis between CATIA and Geant4 COILs∆ ∑/𝑫𝒊𝒇 = ∆ 𝟏 + ∆ 𝟐 + ∆ 𝟑 + ∆ 𝟒 + ∆ 𝟓 + ∆ 𝟔 + ∆ 𝟕 + ∆ 𝟖 = 𝟏 ′ 𝟏𝟑𝟖 𝒌𝒈+𝟏𝟒 𝒌𝒈+𝟏𝟓𝟖 𝒌𝒈+ 𝟏 ′ 𝟕𝟑𝟖 𝒌𝒈−𝟗𝟏𝟏 𝒌𝒈+𝟕𝟕𝟖 𝒌𝒈+ 𝟏 ′ 𝟐𝟒𝟖 𝒌𝒈+ 𝟕 ′ 𝟐𝟗𝟗.𝟓 𝒌𝒈=𝟏 𝟏 ′ 𝟒𝟔𝟐.𝟓𝒌𝒈 CATIA G4

58 IV. Integration conflict checking

59 Integration conflict checking∆ 𝑹𝟏 = 𝑹𝟏 𝑪𝒂𝒕𝒊𝒂 − 𝑹𝟏 𝑮𝒆𝒂𝒏𝒕𝟒 = 𝟗 ′ 𝟓𝟏𝟓 𝒎𝒎− 𝟗 ′ 𝟒𝟖𝟎 𝒎𝒎=𝟑𝟓 𝒎𝒎 ∆ 𝑹𝟐 = 𝑹𝟐 𝑪𝒂𝒕𝒊𝒂 − 𝑹𝟐 𝑮𝒆𝒂𝒏𝒕𝟒 = 𝟓 ′ 𝟐𝟗𝟓 𝒎𝒎− 𝟓 ′ 𝟐𝟕𝟎 𝒎𝒎=𝟐𝟓 𝒎𝒎

60 Conclusion of Hypothesis IIHypothesis #02 has been confirmed: The geometry descriptions in the simulation are not consistent with design geometry description The COIL was divided into 21 separate volume Volume and Weight analyses of the COIL have been implemented; Comparison of the volume and weight between CATIA and XML descriptions have been implemented Important differences have been discovered for the following volumes: Cryostat Bottom missing 1’738 kg., Rib missing 1’248 kg., Thermal Shielding missing 2’020 kg., Inner parts of the COIL missing 5’297.5 kg It is was found that there was not Thermal Shielding in the Geant4 description 11.5 tones missed materials were discovered for Geant 4 geometry 35mm dispositioning of the COIL has been discovered

61 Evaluation of Research

62 Thank you for your attention!Niko Tsutskiridze Georgian Technical University European Organization for Nuclear Research CERN SCCTW’2016, 06/10/2016