predictive maintenance

16
Confidential Saama Technologies, Inc Predictive Maintenance in Refineries An Analytics Approach 5/6/22

Upload: saama

Post on 13-Apr-2017

1.541 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: Predictive Maintenance

ConfidentialSaama Technologies, Inc

May 3, 2023

Predictive Maintenance in RefineriesAn Analytics Approach

Page 2: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Business Challenges

Page 3: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Maintenance – Industry Statistics

46%Of Refinery Shutdowns are for mechanical failures

23%

Of Refinery Shutdowns are for maintenance 92%

Of Refinery Shutdowns are Unplanned

Source : Refinery power failures: causes, costs and solutions - Patrick J Christensen, William H Graf and Thomas W Yeung, Aug 2013

Page 4: Predictive Maintenance

ConfidentialSaama Technologies, Inc

The ChallengeOne of the main challenges oil refineries face is …to maximize asset life span, in the most economical way, while not compromising on safety and reliability

Classic Methods include• Reactive Maintenance• Preventive Maintenance• Condition Based Maintenance

SOURCE: Scanderbeg SauerEnhanced Predictive Maintenance - Pierre Marchand, Oct 31, 2014

Page 5: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Urgent and emergency interruptions to

operations due to equipment breakdowns

Revenue Loss due to Downtime

Inefficient Operations and Supply Chain process

Inefficient Asset utilization

Resource expense for Root cause analysis

Page 6: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Excess Spare parts Inventory

Unnecessary resource Utilization

Opportunity Loss cost of unused maintenance records

High Cost and lower efficiency of Preventive(Unnecessary)

maintenance

SOURCE: Parker Hannifin

Page 7: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Predictive Analytics captures real time equipment data and evaluates historical data to

estimate equipment life cycle for continuous

Equipment Health Monitoring

Page 8: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Predictive Analytics System Advantages

An advanced analytics foundation to

optimize operations planning

Ability to scour past data, identify

patterns & model streaming data

Opportunity to analyze real time monitoring data

Mine Recurring issues, failure indicators & resolutions

A Robust, scalable solution which can integrate with other enterprise systems

i

Page 9: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Application Opportunities

• What equipment to pull in for maintenance & when

• What resources to source & allocate for maintenance

• Birds eye view of real time equipment health

• Measure wear and tear of equipment in its lifetime

• Use Historical data to Identify Leading failure indicators

• Root cause analysis of incident

Day to day maintenance

• What spare parts to keep• Product inventory maintenance

based on upcoming maintenance

Inventory Management

Equipment Health Monitoring

Root Cause Analysis

Operations & Supply Chain• Efficient supply chain

management using predicted maintenance time

• Efficient resource allocation

Page 10: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Return on InvestmentUsing Predictive Maintenance as part of asset management program , a typical 100,000-bpd refinery can have an estimated annual benefits of over USD$3.5M per year.:

• Avoiding abnormal incidents…$500,000 • Reducing lost profit opportunities…$1,750,000 • Reducing maintenance budget…$800,000 • Improving staff productivity…$300,000 • Reducing liability insurance premiums…$200,000

Source : “Quantifying the ROI of an asset performance management program”. Hydrocarbon Processing. T Ayral and M Moran, Meridium, Inc. May 2007.

Page 11: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Predictive Algorithms

Page 12: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Predictive Algorithms for Maintenance

A predictive function based on Current & Historical data used to

derive the measures

Predictive Maintenance

Design

Binary Logistic RegressionMultinomial

Logistic Algorithm

Supervised Learning Models

Explanatory variables

Usage duration

Temperature

Pressure

Flow Rate

Historical sensory data

Forecasting Models

Health Score of Equipment

Triggers Alarm for

maintenance requirement

Usecase : Real wear measure of Equipment

Page 13: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Usecase : Major failure IndicatorsFind patterns in tracking variables correlating to failure

Historical Maintenance

Data

Decision trees

Regression based models

Identify Root Cause

Predict future malfunctions

Usecase : Uptime Time before failureModelling historical data to calculate from streaming data

Lifespan Analysis Model

Pearson Correlation

Identify operating Variables

associated with Lifespan

Estimate Equipment’s remaining lifespan

Explanatory Variables

Analytical Model

Deduction or

Identification

Outcome

Historical & Real - time

Maintenance Data

Predictive Algorithms for Maintenance

Page 14: Predictive Maintenance

ConfidentialSaama Technologies, Inc

Equipment Status Dashboard

Component FRC21

Unit ID 0010021

Location Holder 08

Equipment Age Real Wear Age0

102030405060708090

Percentage Age of Equipment

High Temp Pressure Vibration0

102030405060708090

100

Failure Indicators Component EquipmentNo. FRK03

EquipmentNo FRK05

Equipment No FRK06

Component 1

Component 2

Component 3

Component 4

Component 5

Series1

020406080

Equipment Wear Progress

Hours Under Use 6708

Unit ID 0010021

Installation Date 26-07-2015

No of Components 58

Hours till Failure 1677

Forecasted expiry 15-10-2015

Choose Component

Component Usage StatisticsCalculate Real

wear of equipmentLifetime wear &

Warning Indicators

Single dashboard to report the overall health status for an entire manufacturing unit

Page 15: Predictive Maintenance

ConfidentialSaama Technologies, Inc

ConclusionEnhance Predictive Maintenance by assimilating data1. Real – time Sensor Data2. Maintenance Data3. Historical Data of Equipment

Identify characteristics affecting breakdown before it happens. Enhance failure predictions Reduce unplanned shutdowns Predict when Maintenance is required Ensure Effective and efficient spending on

proactive maintenance Optimize operating conditions to maximize

equipment lifetime & Optimize Supply Chain Processes

Supply & Output

Inventory

Refinery Operations

Predictive Analytics

Page 16: Predictive Maintenance

ConfidentialSaama Technologies, Inc

References

• Enhanced Predictive Maintenance - Pierre Marchand, Oct 31, 2014• Refinery power failures: causes, costs and solutions - Patrick J Christensen, William H

Graf and Thomas W Yeung, Aug 2013• Proactively detect failure patterns to improve asset productivity and product quality -

Predictive Maintenance and Quality, IBM• Quantifying the ROI of an asset performance management program”. Hydrocarbon

Processing. T Ayral and M Moran, Meridium, Inc. May 2007.