LXRP Asset Optimization

Project Overview

Client: Level Crossing Removal Project (LXRP) Victoria
Duration: 2023-2024
Role: Lead Data Scientist
Impact: 20% Asset Optimization, Reduced Carbon Footprint

The Challenge

LXRP needed to optimize their metro rail bridge assets while: - Reducing capital expenditure - Minimizing material usage - Lowering carbon emissions - Maintaining safety standards - Managing 1.5 million potential loading combinations

My Approach

1. Data Extraction & Processing

Challenge: Client data was provided in unstructured image formats
Solution: Developed custom OCR solution in Python

# OCR Pipeline for Material Data Extraction
def extract_material_data(image_path):
    # Custom OCR implementation
    # Extracted strength parameters from technical drawings
    return material_properties

2. Statistical Analysis at Scale

  • Simulated 50 years of bridge loading data
  • Conducted 50 million unique statistical analyses
  • Used extreme value theory for rare event prediction

3. Dimensionality Reduction

Problem: 1.5 million potential rail combinations
Solution: Applied experimental design and clustering

  • Reduced to 21 critical combinations
  • Maintained 99% coverage of risk scenarios
  • Enabled focused resource allocation

Technical Implementation

Probabilistic Performance Prediction

# Simplified performance prediction framework
class BridgePerformanceModel:
    def __init__(self):
        self.load_model = ExtremeValueDistribution()
        self.resistance_model = ProbabilisticModel()

    def predict_reliability(self, time_horizon):
        # Monte Carlo simulation for reliability
        return reliability_index

Optimization Algorithm

  • Multi-objective optimization
  • Constraints: Safety, Cost, Carbon
  • Method: Genetic algorithms + gradient descent

Key Innovations

1. OCR for Engineering Data

  • First application in this context
  • 95% accuracy in parameter extraction
  • Saved weeks of manual data entry

2. Hybrid Statistical Approach

  • Combined Bayesian and frequentist methods
  • Incorporated engineering judgment
  • Validated against historical failures

3. Real-time Decision Support

  • Dashboard for scenario analysis
  • What-if simulations
  • Risk visualization tools

Results & Impact

Quantitative Outcomes

  • 20% reduction in material usage
  • 15% decrease in carbon emissions
  • $XX million in cost savings (confidential)
  • 99% confidence in safety margins

Qualitative Benefits

  • Improved decision-making speed
  • Better stakeholder communication
  • Reusable framework for future projects
  • Enhanced team data literacy

Technologies Used

Python OCR/Tesseract Bayesian Statistics Clustering HPC/Parallel Computing Extreme Value Theory

Lessons & Insights

Technical Lessons

  1. Data Quality: Investing in data extraction upfront paid dividends
  2. Scalability: Parallel processing essential for millions of simulations
  3. Validation: Engineering intuition crucial for model validation

Business Lessons

  1. Stakeholder Engagement: Regular workshops ensured adoption
  2. Incremental Delivery: Phased approach built confidence
  3. Documentation: Comprehensive docs enabled knowledge transfer

Future Directions

This project established a framework now being considered for: - National rail network optimization - Highway bridge management - Port infrastructure planning - Climate resilience assessment

This project showcases my ability to handle complex, large-scale optimization problems while delivering tangible environmental and economic benefits. The combination of innovative data extraction, advanced statistics, and domain expertise enabled LXRP to achieve their sustainability goals without compromising safety.

← Back to Projects