System Level Monitoring for

Time-Varying Conditions

with Applications to Shovels

Nima Yousefi

Supervisor: Prof. Mike Lipsett

August 18, 2015

2

Outline

• Introduction

• Problem overview - Motivations

• Tools and Methods

• Design of the test equipment: Shovel rig

• Soil Property Estimation

• Condition Monitoring and Fault detection

• Conclusion

• Future work

3

Introduction

• Situational AwarenessIdentify, process, and analyze the critical information about the mission

• Equipment that interact with their environment

Loading units, Wind turbine, etc.

• Using sensory data for the assessment of the operating

environment, and equipment condition

Examples: Interaction Force, Energy balance (In/Out)

Introduction

4

Monitoring environment/equipment

• Operating modes (one or multiple states),

• What can change in the equipment or the environment

• What are the critical parameters

• Normal operating envelope for the monitored parameters

• How Internal/external change affects the monitored parameter

• There is a desire to be able to detect the change

• Once change is detected, what corrective course of action

Repair/maintenance Vs. Change of operating mode

Introduction

5

Limitations of current systems

• Condition monitoring in mining equipment have only been

implemented for a limited number of subsystems

- OEM systems for Engine, Lubrication system, electrical, etc.

• Very few studies have investigated structural damage monitoring

- Difficult to monitor

- Some studies on the drive train monitoring – Still in progress

• ‘Time-varying behavior’ and ‘Active ground movements’ limits the

application of conventional vibration monitoring

- Alternative parameters need to be investigated

- Application of advanced data processing methods needs to be studied

Motivation

Motivation

6

Motivation

Main Contributions

• Development of a test rig for the study of the structural

integrity in time-varying systems

- Simulates the behavior of a loading unit: non-stationary behavior, Soil-tool interaction

- Allows multiple structural damage

- Can be used for characterizing the loading area condition

• Using an energy based method (i.e. acceleration signal) for

monitoring the condition of the equipment

- It works on the energy principals: Minimized negative effect of environmental Vib.

- Applicable to equipment with rotating or reciprocal mechanisms

• Development of a property estimation algorithm for characterizing

mechanical properties of the environment

- Improved prediction of the cutting force: better dig planning

- Increased load area situational awareness

Motivation

7

Tools and Methods

Methodology

8

I. Shovel Rig Designo Mechanical design

o Instrumentation and DAQ

9

Test rig design

Test rig design

• Shovel assembly is based on off-set crank-slider

• It is coupled with a gearbox system (drive train)

• Rotating - Reciprocal

• Task-Specific considerations

Offers soil-tool interaction via a slow cutting/pushing action

Allows variable attack angles α, and depth of penetration H

Accommodates a few structural defects

10

Commissioned test rig

Test rig design

11

II. Environmental Monitoringo Soil-Tool Interaction Model

o Medium Property estimation

12

The interaction force applied by the ground engaging

tool can be modeled

Machine-Ground Interaction

Property Estimation

Coulomb’s Earth Pressure Model

13

Soil property estimation

Property Estimation

• One way to estimate soil properties from the

interaction force is using the inverse model

Models are too complex to be analytically inversed

Minimize the error between the predicted and the actual

measured force

Non-unique estimates

• An alternative method is to use force readings and

solve the non-linear interaction model iteratively

14

Cutting force measurements

• Experimental set up 3 Medium types were studied: Glass beads, Play sand, Oil sand

Lab measurement of the friction angles

Soil cutting trials at 70o and 80o attack angle

Cutting forces are measured

- Force pairs are fed into the estimation algorithm

Produce estimates for friction angles and Cohesion

Property Estimation

15

• Used Newton-Raphson for solving the system of

non-linear equations

Estimation results

Property Estimation

70

80

16

III. Condition Monitoringo Energy dissipation and vibration

o Feasibility study for monitoring

o Advanced methods for processing

17

• As the energy flows into mechanical

subsystems, parts of it turns into

energy losses: i.e. heat, vibration

• In rotating and reciprocal systems,

vibration can exist even under normal

conditions

• Geometrical defects influence the transmitted vibration

• Structural faults can cause energy losses too, which can

be traceable applying principles of energy conservation

Energy flow in/out of the system can be monitored by

monitoring the acceleration signal of the slider

Energy dissipation and Vibration

Condition Monitoring

18

Proof of concept

• Numerical model in CAD

Fault free system

Faulty element

• Analyzed in MATLAB: Evaluated the residue

19

Structural Faults

Condition Monitoring

20

Time series

Condition Monitoring

• Slider’s displacement

• Slider’s Acceleration,

• Crank’s rotational speed,

• Total shaking force

Displacement

Acceleration

Angular Velocity

Shaking Force

Normal Fault A Fault B Fault C

21

• Statistical feature extraction

• Dimension reduction using Principal Components Analysis

• Four classes can be separated using a linear classifier

• Acceleration signal carries fault signatures

• How and when does each fault manifest itself?

Time domain Analysis

Condition Monitoring

22

Frequency domain analysis

• Operating data Motor nominal speed constant at 126 rpm (2.1 Hz)

Output shaft rotated at 37.8 rpm (0.63 Hz) [reducer GB: 0.3]

• Non-stationary condition Variable Speed & Variable Loading

Fourier transform is not the proper tool

Condition Monitoring

23

Time-Frequency analysis

• Hilbert-Huang Transform

Empirical Mode Decomposition

Hilbert-Huang Spectrum 𝐻(𝜔, 𝑡) (HHS)

Marginal Hilbert Spectrum ℎ(𝜔) (MHS)

Condition Monitoring

24

Intrinsic Mode Functions (IMF)

Condition Monitoring

Normal Fault in Joint Housing Fault in Connecting rod Fault in Rolling Element

25

Hilbert Spectrum 𝐻(𝜔, 𝑡)

Condition Monitoring

Normal Fault A

Fault B Fault C

26

Marginal Hilbert Spectrum ℎ(𝜔)

2x

5x

8x

1x

1x

2x

4x

5x

3x

𝟓

𝟐x

𝟕

𝟐x

𝟑

𝟐x

Condition Monitoring

27

T-F Analysis Summary

• IMFs allow to locate the exact moments when the fault is

present, and its magnitude

• Fault type affects the time window required for detection

Looking only at the first 10 seconds, Fault A would have been

hidden to both analysis methods

Some faults are present in the entire dataset, while some only

activate under certain conditions

• Some hidden features of the signal can be unveiled

investigating the IMFs of interest with MHS

• T-F methods such as HHT can provide useful information

about the non-stationary systems that complements the

time-domain analysis

TF Summary

28

Conclusions

• Acceleration signal carries enough information for monitoring

the structural integrity of the equipment

• Hilbert-Huang Transform is an effective tool for monitoring

the condition of the non-stationary system of interest

exposed to structural defects

• Ground-tool interaction force can be used for characterizing

the condition of the environment

• System level monitoring studied in this work, presents an

opportunity to expand the application of situational

awareness

Conclusions

29

Future work

• Higher fidelity soil-tool interaction models

• Nonlinear solvers that can handle more parameters

• Perform condition monitoring

Under variable speed

In presence of external loading

Using angular velocity as a monitored parameter

• Use improved decomposition method, i.e. EEMD

• Use additional features

Statistical analysis of the T-F data

Fusion of the features from both T and T-F

• Develop tools and methods for applying the learnings of

this research in industrial application

Future work

30

Acknowledgment

• Dr. Mike Lipsett

• MECE machine shop: Roger Marchand, Bernie

Faulkner, Albert Lee

• Contributions from

James Hendry ………… Design and drawings

Paul Pineda ……………. Material requisition

Amanda Kotchon ……… Geomechanics testing

• Dr. Bob Koch

31

Thank you.