36
Presented by © Copyright 2013 OSIsoft, LLC. Implementation of the PI System to support Predictive Maintenance Diagnostics Across Oil & Gas Assets in Italy Cristina Bottani Marco Piantanida

Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

Embed Size (px)

Citation preview

Page 1: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

Presented by

© Copyr i gh t 2013 OSIso f t , LLC.

Implementation of the

PI System to support

Predictive Maintenance

Diagnostics Across Oil &

Gas Assets in Italy

Cristina Bottani

Marco Piantanida

Page 2: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Agenda

2

• Introduction

• Business Challenge

• The Predictive Monitoring Process

• Role of the PI System

• Applications, Results & Benefits

• Next Steps

• Conclusions

Page 3: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Introduction: Eni – Global and Diversified

3

Page 4: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

The PI System in Italy – 5 Plants and Associated Fields

4

Crotone & Hera Lacinia

Gas Plant

Val d’Agri Oil Center

Oil and Gas Plant

Fano

Gas Plant Trecate

Oil Stabilization Plant

Cervia-K

Gas Plant

Gagliano

Gas Plant

Planned Q4 2013

Page 5: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Business Challenge - The Val d’Agri Field

• Field is operated by Eni Spa

• Discovery well: MA1 in 1988 (Monte Alpi 1)

• Early production phase started in 1993 with MA2

• A total of 33 producer wells were drilled to date – 30 of them are currently tied in

5

0 20 km

Matera

Taranto

500

200

Potenza

VAL D'AGRIVAL D'AGRI

GiacimentoCaldarosa

GiacimentoTrend 1 VdA

0 20 km

Matera

Taranto

500

200

Potenza

VAL D'AGRIVAL D'AGRI

GiacimentoCaldarosa

GiacimentoTrend 1 VdA

Caldarosa Reservoir

Trend 1 VdA Reservoir

Page 6: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Business Challenge - Wells and Pipelines

6

CF3/4

CF1

CF2CF5/8

Volturino 1

AGRI 1 – CF 6/9

S.ELIA – CF 7

ME NW1/2/9ME W1/10/Alli4

ME 1

ME 6/7

MA W1 – ME 4

MA 5

MA 1 N – ME 3

MA 1/2CM2D

MA E1

MA9

MA 4 X MA 3D

MA 6/7/8

Sez.1

Sez.2

Sez.3

Sez.6

Sez.5

Alli 1/3

Alli 2

Sez.4

INNESTO 2

INNESTO 1

IMPIANTO 1

ME 5

IMPIANTO 2

CENTROOLIO

GRUMENTO

NOVA

TRAMUTOLA

PATERNO

MARSICOVETERE

MARSICO

NUOVOCALVELLO

VIGGIANO

MONTEMURRO

CORLETO

PERTICARA

LAURENZANA

Tratti CompletatiTratti da realizzare

Aree Pozzo CompletateAree Pozzo da costruire / allestire

Aree Pozzo in progress

Area Sezionamento Completata

Area Innesto CompletataAree Pozzo:

Tratti di linea:

Pergola 1

Potenziamento Sviluppo

Val d’Agri

COVA

COVA = Oil Center (“Centro Olio Val d’Agri”)

Wells (33)

Gaterhing Systems

& Pipelines

Page 7: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Business Challenge - Val d’Agri – Facility

The Oil Center in Viggiano (Val d’Agri)

has a target nominal capacity of 104 K

barrels/day

In 2012, the overall oil production was

95 K barrel/day from 26 wells,

representing 30% of overall Eni’s in

Italy

7

Page 8: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Business Challenge - Equipment

Main Characteristics of the Plant

4 Oil Trains (water, oil & gas separation, oil stabilization)

Gas Treatment (compression, dehydration, sweetening, advanced treatment of sulfur)

Power Generation and Utilities

Main rotating equipment

Power Generation: 3 Gas Turbines (fully instrumented)

One spare Turbine is kept offline, on rotation

Compression Unit: 7 low pressure + 6 high pressure reciprocating compressors (not

comprehensively instrumented)

One spare low P Compressor and one spare high P Compressor are kept offline, on rotation

8

Page 9: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Main Challenges & Current Maintenance Process

9

Page 10: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

The Predictive Monitoring Concept

Early detection (from several weeks to several months ahead of failure) of equipment problems

More time to plan and react

10

xxx

0%

20%

40%

60%

80%

100%

Eq

uip

men

t C

on

ditio

n

Life

Machinery

protection Alarm

A

Events and minor damage

F

E E

Machinery

protection Trip

T

Functional failure

Normal operation

Condition-Based

Monitoring

Predictive Monitoring

Page 11: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Predictive Monitoring Challenges

Level of instrumentation of the equipment is often non-optimal

Mathematical methods based on First Principle models could be unable to perform due to the lack of key sensors

Availability of high fidelity historical data was an issue

Often limited to few weeks

Complemented with offline measurements sampled sporadically (vibrations, boroscope inspections, infrared thermography, lube oil analysis, etc)

Data driven methods requiring a comprehensive set of high fidelity historical data would not work

The implementation of the PI System allowed to overcome this problem

11

Page 12: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

The PI System in Italy – 5 Plants Instrumentation

12

Crotone & Hera Lacinia

2 Turbo Compressors

5 Reciprocating Compressors

RTAP Unix DCS

SDI eXPert SCADA

630 tags

Val d’Agri

13 Reciprocating Compressors

3 Gas Turbines

ABB Advant OCS Master 300 DCS

ABB 800xA Aspect/Connectivity Server

1650 tags

Fano

2 Turbo Compressors

Moore APACS DCS

450 tags

Trecate

4 Reciprocating Compressors

Moore ICI DCS

InTouch Wonderware

60 tags

Cervia-K

2 Turbo Compressors

ABB Advant 500 DCS

IMS Stations

210 tags

Page 13: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Overall Architecture – An Integration Infrastructure

13

Page 14: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Perspective on the PI System….

• Power of PI System infrastructure

• Flexibility and richness of connectors of the PI System – Even very old DCS systems (dating back to 1991) have been interfaced

14

Page 15: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

HQ Architecture - “More Eyes on PI”

15

Page 16: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Two Roles of the PI System

• Feed a specialized Predictive Analytics Package (Smart Signal)

• Direct diagnosis tools when dealing with failures that are rapidly developing on the equipment – Cannot be detected by the predictive analytics package

– Can be investigated with PI Process Book and PI Coresight

16

Page 17: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Feeding a specialized Predictive

Analytics Package – Operation Concept

17

Page 18: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Feeding a specialized Predictive

Analytics Package – Operation Concept

18

Page 19: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Predictive Monitoring Software: IT Architecture

19

On premise installation of the software.

Page 20: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Direct Investigation for Rapidly Developing Events

20

• Development of a PI AF hierarchy consistent

with the Predictive Monitoring package

• Development of PI Process Book files linked to PI AF

• PI Coresight linked to PI AF

Page 21: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Development of a PI AF hierarchy Consistent

with the Predictive Monitoring package

• Diagnostics Rules within Smart

Signal are targeting different

subsystems of the rotating

equipment

– e.g. the firing of the “Combustion Cold

Spot” Rule will open the graphs of the

“Combustion” “GG Exhaust Gas

Path” section of the Smart Signal GUI

21

Page 22: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Development of a PI AF hierarchy consistent

with the Predictive Monitoring package

22

Subsystems and their attributes within Smart Signal

Page 23: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC. 23

Development of a PI AF hierarchy consistent with the

Predictive Monitoring package

PI AF & PI ProcessBook Hierarchy Representation

Same Subsystems

Same Attributes

Page 24: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Analysis in PI ProcessBook

24

Linked to PI AF through Element Relative displays

Page 25: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Analysis in PI Coresight

25

Page 26: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Predictive Monitoring Results

Approximately 1 new notification every week (on average)

One out of 3 notifications has been actionable: the site has taken

action to fix or mitigate the root cause of the problem

A variety of equipment problems has been detected

26

Page 27: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Successful Catch: Broken Compressor Discharge Valve

27

Date of Notification: Oct 7, 2011

Equipment: [… omissis …]

Type: Equipment/Mechanical

Description: We have seen an increase in the discharge

temperature from 142 to 153 C.

Resolution: Action was taken during shutdown and a

broken valve was replaced.

Insight: Discharge Temperature helps detect valve

issues. Individual temperatures at each cylinder can help

further localize the bad valve.

Page 28: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Successful Catch: Lubrication Problem and prevention of potential wear

28

Date of Notification: Jul 12, 2012

Equipment: [… omissis …]

Type: Equipment/Electrical

Description: The temperature of bearings 1 and 2 has

been increasing up to ~69 C with model prediction at ~58

C.

Resolution: An electrical problem had shut down a

cooling fan at the lube oil cooler. Connection was

repaired.

Insight: Early detection of problem prevents unnecessary

wear on the equipment.

Page 29: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Understand of the dynamics of the failure and help

prevent the occurrence of similar cases in the future

29

Trip of Gas Turbines causing a partial controlled

shutdown of the plant.

Sudden increase in fuel gas due to

sudden increase in the P and incorrect behavior of the

regulation valve Failure Occurring without early warnings

Deeper

Analysis

of

Data

Page 30: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Benefits • The following benefits were achieved:

– Prevented damage on electrical motor windings on 3 compressors - 150 K euro

• The following benefits are expected:

– Increased production due to reduction of unexpected equipment downtime with

consequent unplanned production losses

• The future prevention of a case like the shutdown of the Gas Turbines would avoid a

15,000 barrel of lost production

– Reduced operating expenditure due to unscheduled downtime and minimized

equipment damage

• 1.1% – 2.4% increase the availability of the equipment

– Optimization of maintenance operation

• Expected saving in the range 250 – 400 K euro per year

30

Page 31: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Future Plans and Next Steps

• Other assets are being integrated (Nigeria, Alaska,

Congo, Pakistan)

– PI Web Parts, PI AF

• Centre of Excellence on ESP Pumps Worldwide

– PI Web Parts, PI AF, PI Coresight, mobile devices

31

Page 32: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Solution Results and Benefits

Summary

Business Challenge

• Diverse instrumentation and control

environment

• Lack of high fidelity historical data

• Rotating Equipment is subject to a

variety of issues due to the variable

operating conditions of an Oil &

Gas plant

• Need for proactive analytics

The implementation of the PI System

allowed to support Predictive Diagnostics of

rotating equipment across a number of

assets in Italy

32

• PI System – integration infrastructure

• Retention of high fidelity historical

data

• Integration and alignment of asset

hierarchy with predictive monitoring

package

• Use of PI System analytics,

notification & visualization capabilities

• Better visibility on the activities

of OEM technicians in the plant

• 16+ actionable diagnostics with

avoided equipment damage

• Increased overall equipment

availability

• Reduced maintenance costs

• Culture of continuous

improvement

Page 33: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Conclusions

33

• The PI System allowed to effectively introduce the

Predictive Monitoring concept across several Oil &

Gas plants within Eni

– Flexibility to cope with very old DCS ad SCADA systems

– Capability to feed data to specialized Predictive Analytics

packages

– Richness of graphical tools to perform direct analysis of

data (PI Process Book, PI Coresight)

Page 34: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Cristina Bottani [email protected]

Technical leader TS Services for Operations

eni e&p

34

Marco Piantanida [email protected]

Technical leader TS Services for Development

eni e&p

Page 35: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

© Copyr i gh t 2013 OSIso f t , LLC.

Please don’t forget to……

35

Complete the Online Survey

for this session Share what you saw with

friends on Twitter, Facebook

or Linkedin!

Eventmobi.com/emeauc13

#UC2013

Page 36: Implementation of the PI System to support Predictive ... lube oil analysis, etc) Data driven methods requiring a comprehensive set of high fidelity historical data would not work

Brought to you by

© Copyr i gh t 2013 OSIso f t , LLC.