Generation of virtual manikins from anthropometric data ... final... · Generation of virtual manikins from anthropometric data and digital human modeling tools applications ANNUAL

Generation of virtual manikins from anthropometric data and digital

human modeling tools applications

ANNUAL AUDITION OF PH.D. STUDENTS - XXX CYCLE

Advisor: Prof. Maria Pia Cavatorta Ph.D. student: Raffaele Castellone

POLITECNICO DI TORINODepartment of Mechanical and Aerospace Engineering

12/10/2017

Ph.D. scholarship granted by FCA

ROLE OF VIRTUAL FACTORYIn industrial contexts, it is nowadays binding to run ergonomic assessment in the initial stages of the design and industrialization of new work processes

Digital Human Modelling (DHM)

in industrial contexts Improving productivity and saving cost

Prevention of musculoskeletal disorders

DHM tools differ with each other for:

FIELD OF USE:• Posture prediction• Reachability/accessibility

assessments• Postural comfort

evaluation• Biomechanical analysis

COMPLEXITY AND ACCURACY:

More complex models => Greater simulation time

(especially for preprocessing)

2POLITECNICO DI TORINO - Department of Mechanical and Aerospace Engineering

Optimization of the workplace design

3

HUMAN MODEL (A basic software for the ergonomic assessments of the work stations):• Study, improvements and validation

FLOW CHART OF PH.D. ACTIVITIESVALIDATION

COMPARISON WITH A LITERATURE REFERENCE:• Posture prediction and

body angles differences

• Influence of differences on biomechanical analysis

APPLICATIONS

ANTHROPOMETRIC DIFFERENCES IN DESIGN:• Evaluation of limiting users as

compared to classical P50male manikin

COMPREHENSIVE MAPPINGS:• Generation of normalized

maps of reachability

VIRTUAL MANIKINS GENERATION:• Creating virtual manikins with the updated anthropometric models (different approaches)• Implementation of anthropometric models in DHM commercial software

“LA FABBRICA SI MISURA”:An updated and manikin-orientedanthropometric database

Anthropometric model

• Review of check points and anthropometric measurements• Enlargement of the anthropometric database (other populations and percentiles)

KinematicsStudy of the algorithm for the mannequin’s movements (direct and inverse kinematics)

• Literature review of the kinematics and the possibility to manage more d.o.f.)

• Search for one or more objective functions for more realistic posture prediction

• Inverse kinematics and posture prediction

• Comparisons with 3DSSPP (software developed by University of Michigan)

• Minimizing deviation of the postural angles from neutral position

• Minimizing energy consumption

• Taking into account task characteristics like force exertion

Goals:• To provide guidelines for workplace design• Optimization of the operator’s worktasks(reachability, workspace, postural comfort)

Requirements:• Easy to implement at early design stage• Consistency in ergonomic assessments

POLITECNICO DI TORINO - Department of Mechanical and Aerospace Engineering

HUMAN MODEL (HM)

4

6 d.o.f.

3 d.o.f.

Postural angles:• forward bending of the trunk• frontal elevation of the arm• flexion of the elbow

The kinematic system is “redundant”(for 3 d.o.f. →

∞ possible postures)

Current kinematics routine with kinematic conditions (reduction of 1 d.o.f.)

Kinematic conditions:1. within arm reach,

forward bending of the trunk = 0

2. otherwise, elbow flexion = 0


POSTURE PREDICTION WITH HM

5

• Software developed by University of Michigan

• Based on a postural database of experimental reachability tests

• The biomechanical model is implemented in several commercial software

200 900900

1700

Trunk bending Upper arm horizontal angle Upper arm vertical angle

Forearm horizontal angle Forearm vertical angle

Postural angles:

• 3D arm elevation• Elbow included (ISO 11226,UNI-EN1005-4)


LITERATURE REFERENCE (3DSSPP)

6

θ Y=200 Y=300 Y=400 Y=500 Y=600 Y=700 Y=800 Y=900

Z=1700 0 0 0 0 1 11 24 NR

Z=1550 0 0 0 0 0 5 16 29

Z=1400 0 0 0 0 0 4 14 25

Z=1200 0 0 0 0 0 7 17 26

Z=1100 0 0 0 0 1 11 20 29

Z=900 0 0 0 7 15 23 31 39

HM

Trunk bending (θ)

0°- 20°

20°- 60°

>60°

(ISO 11226,UNI-EN1005-4)

Differences analysis:• HM underestimates θ (with 3DSSPP

more trunk bending for visual needsand force exertion)

• For close workpoints,Z ↓ : 𝜃𝐻𝑀 − 𝜃3𝐷𝑆𝑆𝑃𝑃 ↑

(kinematic condition 1)

• When kinematic condition 1→2 : 𝜃𝐻𝑀 − 𝜃3𝐷𝑆𝑆𝑃𝑃 ↓

(In HM no lateral elevation of the arm)

• 2 out of 48 postures exhibited different traffic light evaluations

• Average error = 6.3°• Standard deviation = 3.3°

*Castellone, R., Sessa, F., Spada S., Cavatorta M.P., “Mappatura di angoli posturali e confronto tra strumenti di Digital Human Modeling per prove di raggiungibilità”, XI Congresso Nazionale SIE, Napoli 16-18 novembre 2016


COMPARISONS HM/3DSSPP

θ 200 300 400 500 600 700 800 900

1700 0 0 1 4 7 15 33 NR

1550 0 1 3 5 7 11 21 39

1400 2 3 5 6 9 12 18 30

1200 5 6 8 10 12 15 21 33

1100 7 8 10 11 14 17 24 36

900 11 12 14 16 19 27 36 45

3 DSSPP

7

HM

α Y=200 Y=300 Y=400 Y=500 Y=600 Y=700 Y=800 Y=900

Z=1700 86 82 84 91 117 129 146 NR

Z=1550 54 54 58 67 82 108 121 138

Z=1400 12 23 35 47 64 93 105 119

Z=1200 EST. EST. 14 31 54 79 91 104

Z=1100 EST. EST. 11 30 64 74 86 99

Z=900 EST. 3 25 48 58 69 81 94

HM

α Y=200 Y=300 Y=400 Y=500 Y=600 Y=700 Y=800 Y=900

Z=1700 90 84 84 89 99 118 146 NR

Z=1550 64 61 64 69 78 92 112 140

Z=1400 32 37 45 52 63 76 94 116

Z=1200 EST. EST. 28 37 48 62 80 98

Z=1100 EST. EST. 24 32 44 58 77 93

Z=900 EST. 13 18 29 46 60 74 94

3 DSSPP

Zone 1: Workpoints beyond arm reachability

Abduzione 200 300 400 500 600 700 800 900

1700 60 47 39 30 18 8 11 NR

1550 62 51 44 37 28 17 12 13

1400 46 45 41 38 31 22 11 6

1200 34 37 37 34 29 21 13 13

1100 32 33 35 32 27 17 9 13

900 26 26 25 20 12 11 12 11

3DSSPP (disallineamento G-S)

Zone 2 : Workpoints within arm reachability

HM

α Y=200 Y=300 Y=400 Y=500 Y=600 Y=700 Y=800 Y=900

Z=1700 86 82 84 91 117 129 146 NR

Z=1550 54 54 58 67 82 108 121 138

Z=1400 12 23 35 47 64 93 105 119

Z=1200 EST. EST. 14 31 54 79 91 104

Z=1100 EST. EST. 11 30 64 74 86 99

Z=900 EST. 3 25 48 58 69 81 94

HM

α Y=200 Y=300 Y=400 Y=500 Y=600 Y=700 Y=800 Y=900

Z=1700 90 84 84 89 99 118 146 NR

Z=1550 64 61 64 69 78 92 112 140

Z=1400 32 37 45 52 63 76 94 116

Z=1200 EST. EST. 28 37 48 62 80 98

Z=1100 EST. EST. 24 32 44 58 77 93

Z=900 EST. 13 18 29 46 60 74 94

3 DSSPP • Zone 1: mean abduction =13°(1 out of 48 postures exhibited different traffic light evaluations)

• Zone 2: mean abduction =33°(5 out of 48 postures exhibited different traffic light evaluations)

3DSSPP 3DSSPP


SHOULDER JOINT

8

9

BIOMECHANICAL ANALYSIS

Torso (Flex/Ext) Shoulder (Abd/Add)

Required Population strength Required Population strength

Work point Posture θ(°) M (Nm) εM (%) Mean SD Cap (%) α (°) M (Nm) εM (%) Mean SD Cap (%)

Z=900,Y=400, X=270 HM 0 50-44%

220 69 99 25 -4359%

77 19 96

Z=900,Y=400, X=270 3DSSPP 14 90 254 80 97 18 -27 72 18 99

Z=1100,Y=600, X=200 HM 1 80-30%

223 70 97 64 -6538%

78 19 74

Z=1100,Y=600, X=200 3DSSPP 14 115 250 79 96 44 -47 71 17 91

Z=1200,Y=600, X=270 HM 0 78-30%

220 69 97 54 -6833%

77 19 69

Z=1200,Y=600, X=270 3DSSPP 12 111 247 78 96 48 -51 70 17 87

Z=1700,Y=400, X=270 HM 0 590%

220 69 98 84 -4612%

71 18 92

Z=1700,Y=400, X=270 3DSSPP 1 59 225 71 99 84 -41 67 16 94

Z=1400,Y=900, X=270 HM 25 187-8%

272 86 84 119 -704%

70 17 49

Z=1400,Y=900, X=270 3DSSPP 30 203 285 90 82 116 -67 72 18 61

• The underestimation of the trunk bending angle causes an increase in the calculated moment of shoulder abduction

• 2D kinematics suits the static strength demand approach used in many DHM software programs (independent axes)

• According to recent studies (La Delfa et al, 2014 e 2016) the maximum force exerted depends on the posture of the arm and the direction of the force and DHM tools could overestimate the maximum force exerted due to the error of posture prediction

*Castellone, R., Sessa, F., Spada, S., & Cavatorta M. P. (under revision). “Reach posture prediction through a simple multibody model for early design checks”, Int. J. Ind. Ergon.

ANTHROPOMETRIC DIFFERENCES IN DESIGN

10

• The design phase is made of several loops due to variations in the product or process• For each design loop, an ergonomic assessment is required• Experimental tests or virtual reproductions of the workstation with DHM are time consuming• At this stage, a software that allows to evaluate the influence of anthropometry in a short time is useful

1. Tubes assembly 2. Antenna assembly 3. Gasket assembly


11

Working Point (mm) Manikin Trunk bending (α)Upper arm

elevation (γ)

1

Y=500;

Z=1400

P50 M 0° 47°

P5 F 0° 91°

P95 M 0° 30°

2

Y=800;

Z=1400

P50 M 14° 105°

P5 F 42° 166°

P95 M 9° 91°


1. Tubes assembly 2. Antenna assembly

CASE STUDIES 1-2

CASE STUDY 3

12

Working Point

(mm)Manikin

Trunk

bending (α)

Upper arm

elevation

(γ)

3

Y=500;

Z=800

P50 M 18° 47°

P5 F 11° 57°

P95 M 22° 45°

*Castellone, R., Spada, S., Caiazzo, G., & Cavatorta, M. P. (2017). Assessment of Anthropometric Differences in the Design of Workstations: Case Studies of an Automotive Assembly Line. International Journal of Applied Engineering Research, 12(14), 4549-4555.


3. Gasket assembly

Objective:To provide a method for a quick ergonomic evaluation of postural angles (trunk bending and upper arm elevation) during reachability operations

Method:Generation of normalized maps of reachability with respect to characteristic anthropometric body dimensions

Results:Postural comfort assessment for different workers once the coordinates of the working points are scaled to account for the worker’s anthropometry

Tools and guidelines that allow early postural checks on potentially critical working points may be of great help to assist the analyst through the dynamic reality of workstation design.

13

COMPREHENSIVE MAPPINGS

*Castellone, R., Sessa, F., Spada, S., & Cavatorta, M. P."Comprehensive Mappings of Postural Angles on a Normalized Plane of Reachability." Advances in Human Factors in Simulation and Modeling: Proceedings of the AHFE 2017 Conference on Human Factors in Simulation and Modeling, July 17-21, 2017, Los Angeles, California, USA. Vol. 591. Springer, 2017.


STEPS 1. Acquisition of the Postural Angles (ISO 11226, UNI-EN1005-4):

2. Extension of the Mappings


α (P50) Y=200 Y=300 Y=400 Y=500 Y=600 Y=700 Y=800

Z=1700 -2 -1 1 4 7 15 33

Z=1550 -1 1 3 5 7 11 21

Z=1400 2 3 5 6 9 12 18

Z=1100 7 8 10 11 14 17 24

Z=900 11 12 14 16 19 27 36

γ (P50) Y=200 Y=300 Y=400 Y=500 Y=600 Y=700 Y=800

Z=1700 90 84 84 89 99 118 146

Z=1550 64 61 64 69 78 92 112

Z=1400 32 37 45 52 63 76 94

Z=1100 26 22 24 32 44 58 77

Z=900 15 13 18 29 46 60 74

3DSSPP was used for the posture prediction of the P5, P50, P95 M (NHANES anthropometric database)

Different number of input nodes and size of the grid:• 3×7 (21 working points)• 4×7 (28 working points)• 5×7 (35 working points)

Different interpolation method:• Linear• Spline

3. Normalization of the Reachability Plane

Mappings of postural angles normalized with respect to characteristic anthropometric body dimensions of the manikin (Full stature and Arm length)

METHOD VALIDATION

A comparison between the angles obtained from the interpolation mappings and the values directly derived from postural simulations was made

Check points different from the initially mapped

grid of working points

The coefficient of determination R2 for:• Different number of input

nodes and size of the grid• Different interpolation

method

3DSSPP postural

simulations

MATLAB Interpolated

values

Postural angle Interpolation method Input (3×7) Input (5×7)

α Linear 0.85 0.97

Spline 0.97 0.98

γ Linear 0.95 0.97

Spline 0.99 0.99

Trunk bending angle (α)when using a linear interpolation (a) and a spline interpolation (b) for an initial input grid 3×7 in size


POSTURAL ANGLES MAPPINGSThe traffic light scheme was used to define different comfort zones in accordance with the international technical standards to ensure a valid and easy-to-interpret support to the analyst

P50 M Upper arm elevation angle (γ) P50 M Trunk bending angle (α)


Z (mm)

Y (mm) Y (mm)

Z (mm)

Acceptable

To be verified

Unacceptable

(ISO 11226,UNI EN 1005-4)

TRUNK BENDING ANGLE - NORMALIZED REACHABILITY PLANEComparison of the trunk bending angle mappings among the three percentiles in the normalized plane respect to Stature and Total arm length

P95M

𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒

𝑌∗

𝑇𝑜𝑡𝑎𝑙 𝑎𝑟𝑚 𝑙𝑒𝑛𝑔𝑡ℎ

𝑌∗


𝑌∗


𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒

P5M P50M

Z=0,6Y=1


TRUNK BENDING ANGLE COMPARISONS - NORMALIZED PLANE

Comparison among the angular values of the three percentiles in the normalized plane:

• Angular values are very similar among the three percentiles, and the difference is always limited within 3°

Working points at fixed heights:

𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒= 0,55 (Hip height)

α

α

α

𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒= 0,82 (Shoulder height)

𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒= 0,97 (Stature)


UPPER ARM ELEVATION ANGLE COMPARISONS - NORMALIZED PLANE

The comparison among the three percentiles of the postural angle in the normalized plane showed no significant differences due to the different anthropometry

𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒= 0,55 (Hip height)

𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒= 0,82 (Shoulder height)

𝑍∗

𝑆𝑡𝑎𝑡𝑢𝑟𝑒= 0,97 (Stature)

γ γ

The results confirm that the mappings of the postural angles can be used for any

given percentile with an acceptable approximation, provided that the working points are normalized with respect to the

manikin’s anthropometry.

γ


𝑌∗


𝑌∗


Project: “La Fabbrica si Misura”

Create a representative anthropometric database of the Italian working population

Some numbers of the project:• 78000 data to be statistically analyzed (not aggregated data)• 13 anthropometric measurements (referred to the

anthropometric points useful for the identification of the manikin’s check-points)

• 6000 subjects estimated for a significant sample

(UNI EN ISO 15535:2006)

Limitations of the current anthropometric database (ISO 7250:2008):• Deprived of some important measurements required to create a virtual manikin• Does not represent the working population • The anthropometric examinations were conducted in 1990 (secular trend)

Creation of anthropometric manikins with a logic different from

“Classical percentile approach”

POLITECNICO DI TORINO - Department of Mechanical and Aerospace Engineering 20

21

- Define the "minimum/optimal" number of anthropometric measurements (for creating virtual manikins consistent with requirements of technical standards on ergonomics risk assessments)

- Define the minimum sample number to ensure a correct estimate of all percentiles (relative error 1%, confidence level 95%, ISO 15535:2012)

- Definition of measurement and data collection protocols (univocally define the measurement procedure, the sequence of anthropometric measurements and data recording)

- Big data - error check (13 anthropometric measurements, 6000 subjects, 13 different plants, need to check the database to highlight and eliminate errors)

- Activity progress and preliminary statistical analysis (Monthly FCA Telepresence, comparisons with other anthropometric database in collaboration with Department of Life Sciences and Systems Biology – Università degli studi di Torino)

- Virtual manikins implementation with updated data in Siemens Jack and Process Simulate (in collaboration with Università della Campania Luigi Vanvitelli)

ACTIVITIES ON THE PROJECT

22

3D BODY SCAN

Integration of anthropometric measurements with 3D Body scan in collaboration with Università degli studi di Torino, Project funded by Regione Piemonte: “Humans”)

• Comparison with traditional anthropometric measurements to ensure quality standards (ISO 20685-2:2015)

• Locations of landmarks as specified by ISO 7250-1 from the surface shape (ISO 20685-2:2015)

• Obtaining the missing anthropometric measurements of the FSM project to create a 3D virtual manikin

MANIKINS GENERATION


Different approaches for the generation of the virtual manikins• CLASSICAL PERCENTILE (all body dimensions belong to the same percentile).

Database can be aggregated

0

100

200

300

400

500

600

700

StatureP5

0

100

200

300

400

500

600

52 62 72 82 92 102 112 122 132 142

WeightP5Manikin P5 Male:

Stature: 1609 mmWeight: 63Kg

• HUMAN SCALE STANDARD (percentile is on stature, other dimensions are mean values for individuals of the given range of stature)Database have to be disaggregated

0

100

200

300

400

500

600

700

1443 1505 1567 1629 1691 1753 1815 1877 1939 2001

Stature

HSS 1542-1642:Stature: 1609 mmWeight: 72Kg

HSS

0

10

20

30

40

50

60

51 59 67 75 83 91 99 107

Weight

P50

23

MULTIVARIATE MANIKINS WITH PCA


• PCA is a method to reduce a multi-dimensional dataset to a more manageable size.• The number of the PC is chosen considering the percentage of explained variance of the

sample• The dataset used in this PCA were reduced to three orthogonal components comprised of a

linear combination of the original body dimensions

Principal component analysis (PCA) was used as a statistical method to obtain subject anthropometries that represent the extreme body sizes and shapes that are realistically proportioned

• The origin of the PCA reference system represents the “average man” and each measured subject is a point in the new Reference System

• The most distant points from the center represent models with unusual anthropometric measures

24

BOUNDARY MANIKINS


• An ellipsoid was created to include and accommodate a percentage of points (90-95%)

• The ellipsoid points represent the extreme manikins that have to be accommodate

• 14 manikins were genereted: 6 intersection points beetwen the ellipsoid and the PC axes, and 8 centers of each quadrant formed by the axes of the Reference System

• The manikins do not only represent smaller and larger persons but represent differentcombinations of body proportions

• Using these manikins as limiting users in design allows to accommodate a greater percentage of the population

25

TRAINING ACTIVITIES AND PUBLICATIONSHard skill courses:• Ergonomics for Manufacturing Systems - Hard skill – II level course – 60h – Politecnico di Torino• Car body design and aerodynamics - Hard skill – II level course – 100h – Politecnico di Torino• Progettazione a crash di strutture di veicoli - Hard skill – III level course – 30h – Politecnico di Torino

Soft skill courses:• Writing Scientific Papers in English - Soft skill – III level course – 15h – Politecnico di Torino• Public speaking - Soft skill – III level course – 5h – Politecnico di Torino• Comunicare la ricerca ai non addetti ai lavori – Soft Skill – III level course – 10 h – Politecnico di Torino• Managing Ph. D. Thesis as a Project – Soft Skill – III level course – 10 h – Politecnico di Torino

Conference papers:• Castellone, R., Sessa, F., Spada, S., & Cavatorta, M. P."Comprehensive Mappings of Postural Angles on a

Normalized Plane of Reachability." Advances in Human Factors in Simulation and Modeling: Proceedings of the AHFE 2017 Conference on Human Factors in Simulation and Modeling, July 17-21, 2017, Los Angeles, California, USA. Vol. 591. Springer, 2017.

• Castellone R., Sessa F., Spada S., & Cavatorta M.P., “Mappatura di angoli posturali e confronto trastrumenti di Digital Human Modeling per prove di raggiungibilità”, XI Congresso Nazionale SIE, Napoli 16-18 Novembre, 2016.

Journal papers:• Castellone, R., Spada, S., Caiazzo, G., & Cavatorta, M. P. (2017). “Assessment of Anthropometric

Differences in the Design of Workstations: Case Studies of an Automotive Assembly Line”, International Journal of Applied Engineering Research, 12(14), 4549-4555.

• Castellone, R., Sessa, F., Spada, S., & Cavatorta M. P. (under revision). “Reach posture prediction through a simple multibody model for early design checks”, Int. J. Ind. Ergon.

POLITECNICO DI TORINO - Department of Mechanical and Aerospace Engineering 26

THANK YOU FOR YOUR ATTENTION

27

Raffaele CastellonePh.D. Student, Politecnico di TorinoEmail: [email protected]


Generation of virtual manikins from anthropometric data ... final... · Generation of virtual manikins from anthropometric data and digital human modeling tools applications ANNUAL

Documents