• RBI Assessment (Pipelines & Piping): Execute comprehensive Risk-Based Inspection assessments for critical energy assets. Perform pipeline sub-segmentation to isolate segments based on damage mechanism susceptibility and define corrosion circuits/loops and inventory groups for facility piping.

• Asset Life Analysis: Conduct Fitness-for-Service (FFS) and Remaining Life Assessments (RLA) to evaluate the structural integrity of aging equipment and optimize inspection intervals.

• Corrosion Mitigation: Design Impressed Current Cathodic Protection (ICCP) systems for buried parallel pipelines, utilizing COMSOL Multiphysics for advanced numerical modelling and interference analysis.

• Standard Compliance: Ensured all composite piping products (FRP/GRP) met 100% compliance with internal company protocols and international engineering standards, including ASTM, ISO, API, and ASME B31.3.

• Dimensional & Material Inspection: Conducted rigorous inspections of product dimensions, material performance, and fabrication procedures, utilizing AutoCAD to verify accuracy against design specifications.

• Root Cause Analysis (RCA): Spearheaded technical investigations into product defects and customer complaints, identifying systemic failures and implementing corrective actions that reduced persistent manufacturing defect rates from 5% to 3% in one month.

• Conducted metallurgy and materials characterization practicum sessions including mechanical testing and microstructure analysis of metals — materials directly relevant to piping and pressure vessel service environments.
• Supervised 20+ students per session through experimental procedures with strict adherence to safety and laboratory protocols.

• ASIIN International Accreditation
• Thesis: Effect of Temperature and Times of Aging on Hardness and Microstructure of Aluminum Alloy 7075
  • Relevant Coursework: Metallurgy, Materials Characterization, Corrosion (API 571), Failure Analysis, Auto CAD/SolidWorks.
• Project: Automated Motion-Detection & Response Prototype (ITB) | May – June 2020
  • Challenge: Addressed chronic vehicle congestion and unauthorized idling at the Dayang Sumbi gate by developing an automated, non-violent means.
  • Technical Action: https://youtu.be/aQW6wBEDbIo
    • System Integration: Engineered a prototype combining computer vision (CV) with embedded systems, featuring an Arduino-controlled 3-axis servo mechanism.
    • Advanced Logic: Developed a real-time motion-tracking algorithm using Processing IDE and pixel-data analysis.
    • Hardware Optimization: Successfully bypassed Arduino hardware limitations by offloading complex computer vision processing to a laptop-microcontroller interface, ensuring high-speed tracking and response.
    • Prototyping: Designed and assembled a 3-axis response mechanism on a custom-built chassis to execute targeted signalling.
  • Impact: Successfully demonstrated a functional application of computer vision to solve urban infrastructure challenges, proving the viability of automated, low-cost solutions for real-time traffic management.

Python | EasyOCR | TrOCR | Deep Learning | VLM

Situation

Risk-Based Inspection (RBI) assessments require engineers to manually review large-format P&ID and PFD documents (typically A0/A1 size) to identify and number all piping equipment, define corrosion circuits, and establish inventory groups. This manual process is cognitively demanding, time-consuming, and highly prone to human error — especially across interconnected, multi-sheet P&ID sets common in large plant facilities.

Task

To address this bottleneck, I set out to develop an intelligent P&ID analysis tool capable of automating the extraction and interpretation of the three fundamental components of any P&ID/PFD: text, symbols, and lines. The end goal was a system that could:

• Automatically identify and locate all equipment and piping tags

• Recognize instrument and equipment symbols to auto-generate inventory groups

• Trace process lines to support corrosion circuit creation

• Serve as a general-purpose P&ID analysis platform beyond RBI use cases

Action

I developed a Python-based application integrating OCR engines (EasyOCR, TrOCR) and deep learning/Visual Language Models (VLM) for symbol recognition. Key engineering challenges and solutions included:

• PDF format limitation — Converted PDF drawings to high-resolution images (PNG) to enable compatibility with OCR and vision models

• Image scale problem — P&ID images often exceed 10,000 px in dimension while tag text occupies only 10–20 px, exceeding model input limits. I designed a sliding window mechanism with overlapping tiles to process large images in smaller, overlapping segments, eliminating blind spots at tile boundaries

• Symbol diversity — P&ID standards include hundreds of symbol variants. I curated a labeled dataset covering primary symbols (valves, flanges, instruments, etc.) and trained a deep learning model for initial recognition, with plans to expand coverage as compute resources allow

Result

• Text extraction module: Successfully deployed — the tool accurately extracts every equipment and piping tag along with its coordinates, enabling instant location of any item within a P&ID without manual searching, making the equipment numbering step in RBI assessments trivial

• Symbol extraction module: Functional for primary symbols (valves, flanges, instruments); model accuracy improvements are ongoing to cover the full symbol library

• Line extraction module: Planned as the next development phase, to be initiated after symbol recognition reaches production-level accuracy

The tool has significantly accelerated the RBI assessment workflow, reducing both the time and cognitive load required during the P&ID review stage — a process that previously required hours of manual interpretation per document.

Root Cause Investigation & Defect Reduction

• Identified and investigated a persistent curing defect on a Glass Reinforced Epoxy (GRE) production line, in which discrete fibre glass sections exhibited incomplete resin curing, while adjacent sections remained unaffected.

• Formulated a technical hypothesis attributing the defect to an imbalanced or non-uniform application of the binding agent (binder) along the fibre glass substrate; designed a Loss on Ignition (LOI) test methodology to quantify binder distribution across fibre glass rolls, which were sectioned into 50 cm specimens for systematic evaluation.

• Proposed and implemented an adjustment to the resin mixing process, elevating the mixing temperature to enhance resin viscosity and substrate wettability; this intervention reduced the monthly defect occurrence rate from 5% to 3% within the trial period, while maintaining full compliance with company product standards.

Quality Inspection Digitalisation Project

• Identified a critical operational gap: all production, quality inspection, and compliance records were maintained in paper-based format, rendering parameter tracking and data-driven decision-making impractical.

• Proposed and led the design and implementation of a digital quality inspection system within the Microsoft Power Platform ecosystem, comprising:

  • A Power Apps mobile application for field inspectors to capture real-time inspection data, including defect details and photographic evidence.

  • A SharePoint List as the primary data repository (selected as a cost-effective alternative to Microsoft Dataverse for the project's early-stage phase).

  • Microsoft Power Query (Excel) for automated data transformation and cleansing of raw inspection records.

  • A Power BI Dashboard for live monitoring and visualization of quality metrics.

  • A Power Automate workflow to programmatically generate and archive standardized inspection reports directly from the Power Apps interface.

• Conducted a one-month pilot trial; the system successfully captured 100% of quality defect records, the dashboard was operational, and the automated report generation function performed without failure.

• Planned subsequent integration with the production division to enable cross-functional, data-driven analysis of process parameter adjustments — however, this phase was not completed due to a transition to a new workplace.

o   Challenge: Addressed chronic vehicle congestion and unauthorized idling at the Dayang Sumbi gate by developing an automated, non-violent means.

o   Technical Action:

§  System Integration: Engineered a prototype combining computer vision (CV) with embedded systems, featuring an Arduino-controlled 3-axis servo mechanism.

§  Advanced Logic: Developed a real-time motion-tracking algorithm using Processing IDE and pixel-data analysis.

§  Hardware Optimization: Successfully bypassed Arduino hardware limitations by offloading complex computer vision processing to a laptop-microcontroller interface, ensuring high-speed tracking and response.

§  Prototyping: Designed and assembled a 3-axis response mechanism on a custom-built chassis to execute targeted signalling.

o   Impact: Successfully demonstrated a functional application of computer vision to solve urban infrastructure challenges, proving the viability of automated, low-cost solutions for real-time traffic management.