Project Overview

Category	Details
Industry	Oil & Gas Terminal Operations
Project Type	Predictive Maintenance & Cost Forecasting
Duration	36-Month Historical Analysis + 12-Month Forecast
Data Sources	CMMS, SAP ERP, SCADA/PI Historian, ATG, HSE
Tools Used	Python, Pandas, NumPy, XGBoost, Power BI
Objective	Forecast maintenance costs and identify major reliability drivers impacting operational expenditure

Business Problem

Maintenance costs were increasing across terminal operations, but management had limited visibility into:

• Future maintenance expenditure
• Major failure modes driving costs
• Asset reliability trends
• Corrective versus preventive maintenance performance
• Long-term budget requirements

As a result, maintenance planning was largely reactive, increasing the risk of unplanned downtime and budget overruns.

Analytical Approach

Activity	Description
Data Integration	Combined CMMS, SAP ERP, SCADA, ATG, and HSE datasets
Data Validation	Verified completeness, duplicates, data types, and quality
Feature Engineering	Created rolling cost trends, corrective ratios, variance metrics, and temporal features
Predictive Modelling	Developed XGBoost forecasting model
Model Validation	Used TimeSeriesSplit cross-validation and hold-out testing
Forecasting	Generated 12-month maintenance cost projections

KPI Summary Dashboard

(Dashboard 2 – KPI Summary Dashboard)

Purpose

This is the executive overview designed for Operations Managers, Maintenance Managers, Asset Managers, and Finance stakeholders.

What the Dashboard Shows

KPI	Value	Business Meaning
Average Monthly Maintenance Cost	$10,989	Average maintenance spending across 36 months
XGBoost R²	0.866	Model explains 86.6% of cost variation
MAE	$706	Forecast error is commercially acceptable
Forecast Total Cost	$178K	Expected maintenance spend next year
Total Work Orders	108	Maintenance workload analyzed
Corrective vs Preventive	67% vs 33%	Reactive maintenance dominates
Budget Overrun	100%	The majority of months exceeded budget, indicating persistent maintenance cost pressure.
Cross Validation Error	$1,425	Forecast remains stable across time periods

---

Business Insights

Insight 1

Maintenance spending consistently exceeds budget expectations.

Insight 2

The terminal relies heavily on corrective maintenance, indicating failures are often addressed after occurrence.

Insight 3

Forecasted maintenance expenditure is significantly higher than historical budget allocation. The model forecasts a continued upward cost trend driven by increasing downtime and corrective maintenance activity, resulting in projected expenditure approximately 35% above the historical average.

Management Recommendation

• Revise maintenance budget for upcoming fiscal year.
• Increase preventive maintenance coverage.
• Establish reliability improvement initiatives for high-risk assets.
• Use predictive maintenance planning rather than reactive maintenance execution.

Historical Maintenance Cost Trend Analysis

(Dashboard 3 – Historical Maintenance Cost Trend)

Purpose

To understand maintenance cost behavior over the past 36 months.

---

What the Dashboard Shows

• Monthly maintenance cost trend
• Historical variability
• Comparison against budget baseline

Business Insights

Observation 1

Maintenance costs fluctuate between approximately:

• Lowest ≈ $6K
• Highest ≈ $18K

indicating unstable maintenance demand.

Observation 2

Several months experienced major cost spikes.

These spikes likely represent:

• Emergency repairs
• Equipment breakdowns
• Unplanned maintenance activities

Observation 3

The historical average remains above budget expectations.

Business Impact

Unpredictable maintenance costs create:

• Budgeting challenges
• Resource allocation issues
• Increased operational risk

Recommendation

Implement reliability-centred maintenance practices to stabilize maintenance expenditure.

Budget vs Actual OPEX Analysis

(Dashboard 5 – Budget vs Actual Comparison)

Purpose

To determine how maintenance spending affects financial performance.

What the Dashboard Shows

Monthly variance between:

• Budget
• Actual OPEX

Business Insights

Observation 1

The terminal consistently operates above budget.

Observation 2

Average overspend:

+$8,000/month

which aligns with earlier ERP findings.

Observation 3

No evidence exists of sustained budget recovery periods.

Business Impact

If the trend continues:

Annual overspend could exceed:

$96,000 annually

Recommendation

Management should:

Immediate

• Re-baseline maintenance budget

Medium-term

• Investigate recurring maintenance drivers

Long-term

• Integrate forecasting outputs into annual budget planning

Section 4: Model Performance Validation

(Dashboard 6 – Model Performance Metrics)

Purpose

To verify whether management can trust the forecast.

What the Dashboard Shows

Metric	Result
R²	86.6%
MAE	$706
RMSE	$1,155
CV MAE	$1,425

Business Insights

Observation 1

The model captures most maintenance cost behaviour.

Observation 2

Forecast errors remain relatively small compared with monthly spending levels.

Observation 3

Cross-validation confirms the model generalizes well.

Business Impact

The forecast is reliable enough to support:

• Maintenance planning
• Budget forecasting
• Resource allocation decisions

Recommendation

Adopt quarterly model refresh cycles using updated CMMS data.

Cost Tier Classification & Risk Analysis

(Dashboard 7 – Confusion Matrix)

Purpose

To evaluate how accurately the model identifies low, medium, and high maintenance cost periods.

What the Dashboard Shows

Overall Accuracy:

77.8%

Business Insights

Observation 1

The model correctly identified:

• All Low-cost months
• All High-cost months

Observation 2

Misclassifications occurred mainly between:

Medium → High

This is operationally safer.

Why This Matters

Underestimating future maintenance costs creates greater business risk than overestimating them.

The model behaves conservatively.

Recommendation

Use the model as an early warning tool for:

• High-cost maintenance periods
• Resource planning
• Contractor scheduling

Failure Mode & Cost Driver Analysis

(Dashboard 8 – Feature Importance Analysis)

Purpose

To identify what drives maintenance cost increases.

What the Dashboard Shows

Top Predictive Drivers

Driver	Importance
Cost per WO	49%
Rolling 3-Month Downtime	20%
Month of Year	17%
Rolling Cost Trend	8%

Business Insights

Key Finding 1

Cost per Work Order contributes almost half of forecast behaviour.

This suggests:

Maintenance quality and repair efficiency directly influence future costs.

Key Finding 2

Downtime history is the second strongest predictor.

Assets experiencing repeated downtime today are likely to generate future maintenance costs.

Key Finding 3

Failure Mode Analysis

Highest Cost Drivers:

Failure Mode	Cost
Valve Leak	$84,465
Alignment Issues	$82,196

Highest Frequency Failure:

Failure	Insight
Bearing Failure	Occurs most frequently

---

Business Impact

Valve leaks and alignment issues consume the largest maintenance budgets.

Bearing failures drive recurring maintenance workload.

Recommendation

Reliability Program

Target:

• Valve leaks
• Alignment failures

Potential savings:

≈ $33,000 annually, estimated savings based on scenario modelling

Condition Monitoring

Deploy:

• Vibration monitoring
• Bearing health monitoring
• Predictive diagnostics

using SCADA/PI Historian infrastructure.

12-Month Maintenance Cost Forecast

(Dashboard 4 + Dashboard 9)

Purpose

To estimate future maintenance expenditure and support annual planning.

What the Dashboard Shows

Forecast range:

Month	Forecast
Lowest	~$14K
Highest	~$16K–17K

Forecast Total:

$178,000

Business Insights

Observation 1

Forecasted costs exceed historical budget assumptions.

Observation 2

All forecast months fall into:

High Cost Tier

This indicates persistent maintenance pressure.

Observation 3

Forecast confidence intervals remain relatively stable, suggesting moderate forecasting uncertainty.

The model expects continued cost growth rather than stabilization.

Business Impact

Current budget levels will likely be insufficient.

Without intervention:

• Maintenance costs continue increasing
• Reliability deteriorates
• Financial pressure rises

Assumptions & Limitations

Assumptions

• Historical patterns continue into forecast period
• Work order quality remains consistent
• No major asset replacement occurs

Limitations

• Dataset contains 36 months of history
• Failure mode labels were standardized
• External operational disruptions not included

Recommendation Summary

Priority	Action	Expected Outcome
High	Increase annual maintenance budget from $100K to ~$180K–$210K	Realistic financial planning
High	Launch Reliability-Centred Maintenance (RCM) programme	Reduce major failure modes
High	Implement vibration-based monitoring for pumps and rotating equipment	Reduce bearing failures
Medium	Shift maintenance strategy from 67/33 to 50/50 preventive ratio	Reduce unplanned downtime
Medium	Integrate SCADA condition data with CMMS planning	Enable predictive maintenance
Medium	Refresh forecast quarterly	Maintain forecast accuracy

Conclusion

The analysis demonstrates that maintenance costs are being driven primarily by recurring equipment reliability issues rather than isolated incidents. Historical spending patterns show consistent budget overruns, while predictive modelling indicates maintenance expenditure will continue rising over the next 12 months. Valve leaks, alignment issues, and recurring bearing failures represent the most significant opportunities for cost reduction and reliability improvement.

By combining CMMS, SCADA, SAP ERP, HSE, and operational data into a unified analytics framework, management gains visibility into maintenance performance, future spending requirements, and emerging asset risks. The recommended strategy is to move from a reactive maintenance model toward a reliability-centred and predictive maintenance approach, supported by condition monitoring, improved planning, and data-driven budgeting. This transition can improve asset availability, reduce downtime, control maintenance costs, and strengthen long-term operational performance

Hire Me to Build Operational Analytics & Predictive Maintenance Solutions

Are your operational, maintenance, HSE, inventory, or ERP datasets sitting in separate systems with limited visibility?

I help organizations transform operational data into actionable business intelligence by integrating systems such as:

• SCADA / PI Historian
• IBM Maximo / CMMS
• SAP ERP
• ATG Inventory Systems
• HSE Management Systems
• Excel, SQL & Data Warehouses

Services

✔ Operations Performance Dashboards
✔ Maintenance Cost Forecasting
✔ Predictive Maintenance Analytics
✔ Asset Reliability & Downtime Analysis
✔ OPEX & Budget Variance Analysis
✔ SAP & CMMS Reporting Solutions
✔ KPI Design & Executive Reporting
✔ Power BI Dashboard Development
✔ Data Cleaning, ETL & Data Integration
✔ Business Intelligence & Decision Support

Business Outcomes

• Reduce unplanned downtime
• Improve asset reliability
• Optimize maintenance spending
• Increase visibility across operations
• Improve budget planning accuracy
• Enable proactive decision-making

Industries

• Oil & Gas
• Energy & Utilities
• Manufacturing
• Healthcare
• Logistics & Supply Chain
• Education & Public Sector

Let’s Discuss Your Project

Whether you need a Power BI dashboard, predictive maintenance model, operational KPI framework, or an end-to-end analytics solution, I can help transform your data into measurable business value.

📧 Email: brightegho700@gmail.com
💼 LinkedIn: linkedin.com/in/bright-egho-108305280

Available for freelance, contract, remote, and full-time opportunities.