Skip to main content

Sho Okazaki

Research Projects

My research connects engineering knowledge with data-driven learning across railway infrastructure and manufacturing systems.

Railway asset management Manufacturing anomaly diagnosis

Active

UCL

University of Cambridge

Whole-System Approach to Railway Asset Management

Models railway drainage and track as an interdependent system to improve risk assessment and maintenance planning.

Research question

How can drainage and track be assessed as one interdependent railway system?

Approach

Combines machine learning, hierarchical Bayesian modelling, and interdependency analysis to link drainage condition with track deterioration.

Machine learning

Hierarchical Bayesian modelling

Asset interdependency analysis

How the research developed

2026
Journal of Infrastructure Systems
Machine Learning Approach to Redefining Risk in Railway Drainage Systems
Reframed drainage risk with data-driven learning methods.
1. 2025
  Reliability Engineering & System Safety
  Railway track performance prediction considering track-drainage interdependencies
  Extended the track deterioration model to consider interdependencies with drainage assets.
2. 2026
  ICONHIC
  A Hierarchical Bayesian Network for Modelling Multi Component Drainage Effects on Railway Track Deterioration
  Established a probabilistic view of component-level drainage-driven deterioration.

Completed

University of Tokyo

Knowledge-Driven Anomaly Detection and Diagnosis in Manufacturing Systems

Combines engineering knowledge and operational data to detect anomalies and identify their causes in manufacturing systems.

Research question

How can engineering knowledge and operational data be combined to detect anomalies and identify their causes?

Approach

Links FMEA, maintenance records, knowledge graphs, language models, and multimodal sensing across detection and diagnosis.

Knowledge graphs

Graph learning

Multimodal anomaly detection

How the research developed

2023
2023 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM)
A framework to support failure cause identification in manufacturing systems through generalization of past FMEAs
Established fault-cause diagnosis by combining fault knowledge and structural knowledge.
1. 2025
  arXiv preprint
  Fault Cause Identification across Manufacturing Lines through Ontology-Guided and Process-Aware FMEA Graph Learning with LLMs
  Expanded inference with LLMs and graph learning.
  1. 2026
    Computers in Industry
    A spatio-temporal anomaly detection system to support understanding of abnormal phenomena in automated manufacturing lines
    Moved the pipeline upstream to anomaly detection before diagnosis.
2. 2024
  2024 IEEE International Conference on Advanced Intelligent Mechatronics (AIM)
  Description method and failure ontology for utilizing maintenance logs with FMEA in failure cause inference of manufacturing systems
  Extended the diagnosis workflow with maintenance records and failure ontology.
  1. 2026
    The International Journal of Advanced Manufacturing Technology
    FBS model-based maintenance record accumulation for failure-cause inference in manufacturing systems
    Introduced FBS-based data input and record accumulation.