Adding Sensor Diagnostics into a Prognostics Model to Better Predict End of Life Gearbox Projections
October 15, 2014
Model-based, multi-physics based prognostics computational technologies
and services
Our applications help extend the remaining useful life (RUL) of new and existing
mechanical systems
The newest prognostics health management (PHM) application for
condition-based maintenance (CBM)
Sentient Science is Based on Three Fundamental Capabilities
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Our 10 Year Research Pedigree Invited a New way to Measure and Test Rotating Equipment Computationally
DEPARTMENT OF DEFENSE
DEPARTMENT OF ENERGY
NATIONAL SCIENCE FOUNDATION
Adding Sensor Diagnostics Into a Prognostics ModelApril 15, 2023
Sentient Science Family
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Sentient Science InnovationExtend the remaining useful life (RUL) through prognostics
Use sensors predicatively rather than diagnostically to operational failure
Sentient Science InnovationExtend the remaining useful life (RUL) through prognostics
Use sensors predicatively rather than diagnostically to operational failure
B – Crack Initiation
Gap between current diagnostic state D and future prognostic state
C.
Extended Life through
prognostics
A – Asset enters service
D – Operational Failure
C - Failure can be confidentlypredicted - PROGNOSIS
®
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
DigitalClone®
Nucleation & PropagationToo Late
Adding Sensor Diagnostics Into a Prognostics ModelApril 15, 2023
Sentient’s Prognostics + Sensor Diagnostics = Model Data Fusion
Technical Approach and Advantages
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Definitions
• CBM Value Statement: Tells you what turbine has started to operationally fail with a few months warning.
– Sensors used to provide additional condition status updates on top of control system sensors
• Prognostics (PHM) Value Statement: What can you do now to keep it from failing?
– Only already installed control system sensors are needed to know failure risk levels
– CBM system sensors are used to identify which turbine at risk will fail
• What is CBM?– Tool that provides feedback on the condition status of a system/component– Typically done through vibration analysis (past 60+ years) or oil debris
sensors (past 50+ years)– Allows you to know 1-6 months advanced notice of failure
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Current State of Vibration Analysis
• High Cost (Upfront, High Number of Sensors)– Number of required sensors unknown– “Over-Sensored”– No upfront optimization
• High Amounts of Data Required (How do you identify a failure?)– Need to see failures with your system– Massive amounts of data before identification of failure– End up with too little or too much data
• Large Amounts of Time Dedicated To Monitoring– Hard to automate until failures are seen in data– Hard to set standard alarm levels, not all “identical” failures look
the same
+ ???
w/o Prognostics
w/ Prognostics
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
1. To provide the same or better accuracy on predictions of component or system failures on a gearbox than traditional CBM alone
• Better interpretation of the sensor signal• Better serialized DigitalClone model for a specific fielded asset
2. To use sensors predictively rather than diagnostically• Using our sensor to validate our prediction for a serialized
asset and to know if we need to recalculate our predictions based on operating conditions
• Sensor is used as an extension to our data model
3. To lower the cost of the sensor infrastructure• Data model indicates the location of potential failure so that we
can target the number and position of the sensors on the gearbox
Sentient’s Prognostics + Sensors
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
The Conventional approach assumes an ideal system with precisely defined parameters which determines the value of the system level output.
Physics-based
Model: ẋ=f(x,p)
Parameters
External ForcesOutput
Prognostics vs. Diagnostics - only Approach
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Sentient innovative approach based on Model Data Fusion (MDF) for better serialized operational failure prediction
- The MDF approach considers the manufacturing tolerances, and reduces the uncertainty in parameters by fusing measurements with a dynamic model. - Decreases the uncertainty bounds in the life estimation of a
component
Physics-based Model: ẋ=f(x,p)
Parameters
External Forces
OutputParameter and Force Estimator
Initial values
Measurem
ents
Prognostics vs. Diagnostics - only Approach
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Prognostics + Diagnostics Red LevelDefinition
Level Operational RUL DefinitionRed in DCL No Change, 12 month or less
RUL, Sensor plus prognostic model
Red Severity 1 12 Months Least severe critical forces seen - Excitation forces seen 1.16 x 104 N (Newton’s) above the 8.0 x 104 N level
Red Severity 2 8 Months 2nd Least severe critical forces seen - Excitation forces seen 2.32 x 104 N above the 8.0 x 104
N level
Red Severity 3 3 Months 2nd Most severe critical forces seen -Excitation forces seen 2.90 x 104 N above the 8.0 x 104 N level
Red Severity 4 1.5 Months Most severe critical forces seen - Excitation forces seen 3.50 x 104
N above the 8.0 x 104 N level
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
DigitalClone LIVE
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Vibration Reports
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Sentient’s Prognostics + Sensor Model = Model Data Fusion
Applications, Demonstration – Wind Examples
Gearbox Operational FailureUptower ReplacementsDerate/Uprate Effects
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Prognostics + Sensors Gearbox Failure Prediction
0 5 10 15 20 25 30 35 400
1
2
3
4
5
6
7
8
9
10x 10
4
Mean
Value
Alert to consistent jump in excitation force
Condit
ion
Indic
ato
r
December, 2014 SeptemberAugust, 2014
DigitalClone identified gearbox failure months in
advance
Magnitude of excitation force due to a spalled intermediate pinion tooth
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
14.8 15 15.2 15.4 15.6 15.8 16
3
4
5
6
7
8
9
x 104
Damaged Gearbox
Healthy Gearbox
SPEED (RPM)
Cond
ition
Indi
cato
r
14.8 15 15.2 15.4 15.6 15.8
2
3
4
5
6
7
8
9
x 104
SPEED (RPM)
Damaged Gearbox
Healthy Gearbox
Cond
ition
Indi
cato
r
Sentient DigitalClone sensor accurately estimated the health state of a faulted gearbox and healthy gearbox
Prognostics + Sensors Gearbox Failure Prediction
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Turbine 01 April 2014 - High Risk Turbine, red alert
• Critical components: low speed pinion, cylindrical bearing
Turbine 02April 2014 - High Risk Turbine , red alert
• Critical components: low speed pinion
Turbine 01 June 2014 (1.0 MW) - Low Risk Turbine , yellow alert
• Critical components: LSP• OVERALL IMPROVEMENT IN GEARBOX OPERATION
July 2014 (1.75MW) - High Risk Turbine , orange alert• Critical components: Bull Gear, Int. LSP, MSBRG• CONTINUING DAMAGE PROGRESSION SEEN
Turbine 02 July 2014 (1.5MW) - Low Risk Turbine and a yellow alert
• Critical components: Int. LSP, MSBRG, ISBRG• OVERALL IMPROVEMENT IN GEARBOX OPERATION
BEFOREREPLACEMENT
AFTERREPLACEMENT
Prognostics + Sensors Gearbox Uptower Replacement, Derate/Uprate Effect
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Prognostics + Sensors Gearbox Uptower Replacement (BEFORE REPLACEMENT)
• Red Alert: High Risk Turbine• Sensor simulations show continuing high risk, especially the intermediate LSP, Downwind MSBRG, and Upwind ISBRG• If CI values, high excitation forces, are consistently much above 8.0e+04 – Red Alert
14.9 15 15.1 15.2 15.3 15.4 15.5 15.6 15.7 15.89
9.5
10
10.5x 10
4
14.9 15 15.1 15.2 15.3 15.4 15.5 15.6 15.7 15.89
9.5
10
10.5x 10
4
Downwind MSBRG
Upwind ISBRG
Red Alert
Red Alert
Simulation Dates: April 2014Simulation Dates: April 2014
14.9 15 15.1 15.2 15.3 15.4 15.5 15.6 15.7 15.86
7
8
9
10x 10
4 CI10 vs. Speed
14.9 15 15.1 15.2 15.3 15.4 15.5 15.6 15.7 15.80
5
10x 10
4 CI12 vs. Speed
Bull Gear
Intermediate Pinion (LSP)
Red Alert
Red Alert
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Simulation Dates: Late June 2014After bearing replacement
10.95 11 11.05 11.1 11.153.5
4
4.5
5x 10
4
10.95 11 11.05 11.1 11.153.5
4
4.5
5x 10
4
Downwind MSBRG
Upwind ISBRG
Yellow Alert
Yellow Alert
• Yellow Alert: Low Risk Turbine• Sensor simulations show new low risk in the Gears & Bearings, with slight elevations in the intermediate LSP• If CI values, excitation forces, are consistently much above 4.5e+04 – Yellow Alert• Improved linearity and consistency seen, indicative of healthier operation.
Simulation Dates: Late June 2014After bearing replacement
10.8 10.85 10.9 10.95 11 11.05 11.1 11.154.75
4.8
4.85
4.9
4.95x 10
4 CI10 vs. Speed
10.8 10.85 10.9 10.95 11 11.05 11.1 11.154
4.5
5
5.5
6x 10
4 CI12 vs. Speed
Bull Gear
Intermediate Pinion (LSP)
Orange Alert
Yellow Alert
Elevated forces at higher speeds
Overall improvement in gearbox operation
Prognostics + Sensors Gearbox Uptower Replacement (AFTER REPLACEMENT)
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Sentient’s Prognostics + Sensor ModelBenefits Summary
• If you already have CBM, Prognostics will provide asset life assessments 1-10 years going forward
• If you do not have CBM today, Model Data fusion between the prognostics (multi-physics model) and the CBM sensor (oil, vibration, other) will provide a highly accurate solution for 1-12 month operational failure
• Prognostics tell you what component will fail and when. The sensor (using MDF approach) is used to confirm predictions instead of waiting for imminent failure
April 15, 2023
Adding Sensor Diagnostics Into a Prognostics Model
Adding Sensor Diagnostics into a Prognostics Model to Better Predict End of Life Gearbox Projections
October 15, 2014