Modern manufacturing depends on supply networks so long and so specialized that few companies can see all the way through them. Nadia Islam Tanha, a Bangladeshi researcher pursuing a Doctor of Engineering in Industrial Engineering at Lamar University in Beaumont, Texas, is working on a question that follows from this: can the failure of such a network be predicted before it happens, rather than explained afterward?
Where the Chain Goes Dark
Her doctoral research focuses on semiconductor supply chains, which stretch from raw wafer materials through fabrication, substrates, and assembly and packaging operations spread across many countries. Most manufacturers know their direct suppliers well. Below that, visibility thins quickly. A specialty chemical produced at a small number of plants, or a packaging facility quietly serving dozens of competing customers, can sit two or three steps down the chain and attract no attention until it stops. Tanha’s work begins from the position that these hidden dependencies are measurable, not merely unfortunate.
Machine Learning as an Early Warning System
The analytical core of her research applies supervised machine learning to signals that manufacturers already generate but rarely combine: supplier delivery performance, lead-time variation, logistics and port-throughput records, and public data on production capacity. Methods such as random forests and gradient-boosted decision trees are suited to this kind of problem, where the warning signs are scattered across many weak indicators rather than concentrated in one obvious metric. Alongside these, she is applying graph and network analysis to map dependency structures and identify which suppliers, materials, or facilities carry disproportionate weight — the nodes whose failure propagates furthest.
Measuring What Recovery Actually Costs
A second strand addresses a question managers face only after a disruption: which response would have helped. Holding buffer inventory, qualifying a second supplier, and reserving surge capacity all carry real cost, and evidence comparing them under different conditions is thin. Tanha’s research aims to produce metrics for how far a disruption spreads, how long recovery takes, and how those outcomes change under each strategy — turning a judgment call into a comparison.
Openness as a Design Choice
Tanha intends the results to be usable beyond any single company. Her plan calls for published findings, documented model structures, and workflows that other researchers can run against their own data. Her earlier peer-reviewed work on manufacturing efficiency, supply chain optimization, and blockchain-based traceability supplies the methodological grounding; she has also served as a peer reviewer for an academic journal.
In her own words: “The most useful thing this research can do is turn a judgment call into a comparison. People choosing between buffer stock, a second supplier, or reserved capacity are mostly working from instinct. I want to give them evidence.”
