Maintenance Philosophies Architecture For Condition Based Maintenance Methods For Fault

Outline Condition Monitoring and Fault Diagnostics Methods and Architectures

Marko Mattila, 2005

AutomationIT

●

Maintenance philosophies

●

Architecture for condition based maintenance

●

Methods for Fault Detection and Isolation (FDI)

●

Examples

Marko Mattila, 2005

Maintenance Philosophies

AutomationIT

Corrective Maintenance

●

Corrective maintenance

●

Run to failure

●

Preventive maintenance

●

The maintenance is done after some part breaks.

●

Condition based maintenance

●

Maintenance costs are low.

●

Marko Mattila, 2005

AutomationIT

Operating costs are high because of the downtime and damages.

Marko Mattila, 2005

AutomationIT

Preventive Maintenance ● ●

●

●

Condition Based Maintenance

Scheduled maintenance action. Schedule is created from lifetime statistics of similar machines. High maintenance costs due unnecessary maintenance actions. Possibility to equipment failures.

Marko Mattila, 2005

AutomationIT

●

Maintenance when required

●

Unnecessary maintenance is avoided.

●

Availability of the equipment is guaranteed

●

Overall cost of maintenance is reduced

●

Monitoring creates costs.

Marko Mattila, 2005

AutomationIT

Architecture for condition based maintenance

Total Cost of Maintenance ●

●

Open System Architecture for Condition Based Maintenance (OSA-CBM) ISO 13374-1&2 Condition monitoring and diagnostics of machines – Data processing, communication and presentation

Lebold et al. 2003]

Marko Mattila, 2005

AutomationIT

Marko Mattila, 2005

AutomationIT

MIMOSA & OSA-CBM ●

Machinery Information Management Open Systems Alliance (MIMOSA) –

●

OSA-CBM ●

Database schema to define all information needed in maintenance management (Common Relational Information Schema, CRIS)

●

MIMOSA OpenO&M standards: – –

OSA-EAI for ERP and operations management defines interfaces to CRIS repositories OSA-CBM for software component interfaces, behavior and structure of their data

Marko Mattila, 2005

●

AutomationIT

Industry led team developed and demonstrated Open System Architecture for Condition Based Maintenance. Boeing, Caterpillar, Rockwell Automation, Rockwell Scientific, Newport news, Oceana Sensor technologies. Partly funded by U.S. Navy Active development from 1999 to 2003.

Marko Mattila, 2005

OSA-CBM Architecture

ISO 13374

• Functional CBM components • Component behavior • Standard interfaces (AIDL) – – – –

●

●

Request Data Get Data Get Explanation Get Config

●

• Data structures

Marko Mattila, 2005

AutomationIT

AutomationIT

ISO Standard Prepared by technical Committee ISO/TC 108, mechanical vibration and shock, Subcommittee SC5, Condition monitotoring and diagnostics of machines. ISO 13374 consist of four parts –

Part 1: General guidelines (ready)

–

Part 2: Data-processing requirements (draft)

–

Part 3: Communication requirements (?)

–

Part 4: Presentation requirements (?)

Marko Mattila, 2005

AutomationIT

ISO 13374 Architecture

Fault Detection and Isolation (FDI)

[Chieng et al, Data-driven techniques for fault detection and diagnosis in chemical processes]

Marko Mattila, 2005

AutomationIT

Marko Mattila, 2005

FDI methods ●

– ●

Based on data measured from the model Statistical Process Control (SPC), Principal Component Analysis (PCA), Artificial Neural Networks (ANN)

●

Straddle Carrier, Research Project

●

X-ray Analyzer, Research Project

Analytical – –

●

Examples

Data-driven –

AutomationIT

Based on detailed mathematical models Parameter estimation, Observers, Parity relations

Knowledge-based – –

Based on qualitative models Causal analysis, Signed directed graph, Expert systems, Case based reasoning, ANN

Marko Mattila, 2005

AutomationIT

Marko Mattila, 2005

AutomationIT

Straddle Carrier

Brakes

• Manufactured by Kalmar Industries • Height 16 meters • Weight 68 tons • 20 or 40 feet containers, weight up to 50 tons

Marko Mattila, 2005

OK

AutomationIT

Marko Mattila, 2005

Difference of desired and measured velocity

Marko Mattila, 2005

Faulty

AutomationIT

Averages of velocity differences

AutomationIT

Marko Mattila, 2005

AutomationIT

Signed directed graph of the breaking system

Courier 3SL X-ray Analyzer ●

●

●

Marko Mattila, 2005

AutomationIT

On-line X-ray fluorescence analyzer for concentrator process Takes samples from up to three sample lines Analyses the concentrations of the elements in the sample slurry

Marko Mattila, 2005

Calibration Validity

Monitoring of Calibration Validity ●

●

●

●

Quality of the analyses is limited by the assay calculation model calibration X-ray measurement data is correlated with the laboratory analyses MLR (Multiple linear regression) model extended with heuristic rules

AutomationIT

●

Ore is not traceable

●

Indicator has to be data-based

●

PCA indicator –

PCA model from the calibration data

–

T2 and Q statistics

Model must be maintained over time

Marko Mattila, 2005

AutomationIT

Marko Mattila, 2005

AutomationIT

PCA Indicator ●

●

PCA Indicator

T2 statistics measure the variation within the PCA model

50

-50 10/01/00 50

Q statistics measures how well each sample conforms the PCA model. –

Cu error% Mean

0 10/21/00

11/10/00

11/30/00

12/20/00

Predicts the assay calculation error

10/21/00

11/10/00

11/30/00

12/20/00

Reacts when ore drifts outside the calibration time PCA model

0 10/01/00 50

01/29/01

10/21/00

11/10/00

11/30/00

12/20/00

01/09/01

01/29/01

T2 statistics 95% limit 0 10/01/00

Marko Mattila, 2005

01/09/01

Q statistics 95% limit

5

–

01/29/01

Zn error% Mean

0 -50 10/01/00 10

01/09/01

AutomationIT

Marko Mattila, 2005

10/21/00

11/10/00

11/30/00

12/20/00

01/09/01

01/29/01

AutomationIT