About Digital Trinity
Digital Trinity is a dynamic team of Master of Science candidates in Industry 4.0 in NUS, collaborating as part of their IND5005B consulting project course. Focused on aiding Micron's digital evolution, the group is specifically tackling Proactive Supply Chain Risk Management: Leveraging AI for Predictive Weeks of Supply and Recovery Insights. By combining their diverse expertise in data science, engineering, and digital business strategy, Digital Trinity aims to deliver innovative, data-driven solutions that enhance Micron's operational resilience, predictive capabilities, and overall efficiency within its complex supply chain.
Meet the Team
Company Overview
Micron Technology, founded in 1978 in Boise, Idaho, has grown into a leading global manufacturer of memory and storage solutions, specializing in critical technologies like DRAM and NAND flash, with an extensive workforce of 43,000 employees across 17 countries. Their primary products, including DRAM (Dynamic Random Access Memory) and NAND Flash Memory SSDs (Solid-State Drives), are integral to a wide range of applications, powering consumer electronics, data centres, mobile devices, and automotive systems. Micron's robust market presence is demonstrated by its global operations, with major manufacturing, R&D, and sales facilities strategically located in the U.S., Taiwan, Singapore, Japan, and India, complemented by significant engineering, sales, and testing operations throughout Europe, China, and Malaysia.
Project Overview
This project tackles the challenge of Micron's "Weeks of Supply" (WoS) metric, which currently offers descriptive insights into raw material inventory rather than proactive guidance for optimal levels amidst demand fluctuations. The project's scope includes simulating ideal inventory levels using Monte Carlo methods to mitigate stockouts and overstocking, optimizing order lead times through historical data analysis, and leveraging machine learning and time series models to predict WoS for proactive procurement, with a prototype developed in Streamlit and Plotly, eventually scaled to a Next.js application. Ultimately, the objective is to construct a data-driven simulation model for Micron's supply chain, utilizing historical data, ARIMA forecasting, Monte Carlo simulations, and reinforcement learning, culminating in an AI-powered dashboard that enables stakeholders to visualize demand and consumption patterns, facilitate inventory optimization, and make more informed strategic decisions.
Feature Overview
Waterfall Analysis
This feature provides a detailed week-by-week breakdown of inventory changes, comparing planned supply and demand against actual purchase order receipts and consumption. It helps identify discrepancies, root causes for inventory imbalances (like inadequate PO coverage or demand spikes), and projects future inventory levels based on confirmed POs.
Material Consumption Analysis
This tool analyzes how materials are used over time, offering insights into overall consumption trends, identifying unusual spikes or drops (outliers), and allowing deep dives into specific materials. It can break down consumption by vendor, plant, and site, helping to understand usage patterns across different dimensions.
Order Placement Analysis
This feature examines patterns in how purchase orders are created. It visualizes overall ordering trends, allows for analysis of specific materials, and can break down order quantities by supplier and plant. It also includes an ABC analysis to categorize materials based on their order value.
Goods Receipt Analysis
This focuses on the process of receiving materials, tracking quantities received over time. It highlights overall goods receipt patterns, detects outliers, and enables detailed analysis for specific materials, including trends and receipts by plant, site, and vendor.
Lead Time Analysis
This feature measures and analyzes the duration between placing an order and receiving the goods. It processes order placement and goods receipt data to calculate actual lead times, compares them against planned lead times, and provides summaries at both material and supplier levels to identify delays or inconsistencies.
Forecast
This capability predicts future material demand based on historical consumption data. Users can select different forecasting models (like XGBoost or ARIMA), specify the forecast period, and account for seasonality to generate demand projections for specific materials.
Inventory Simulation
This tool runs Monte Carlo simulations to model inventory levels under various scenarios. It considers inputs like initial stock, reorder points, lead times, and demand variability (either fixed or based on statistical distributions). The simulation compares reactive versus proactive ordering strategies, highlighting potential stockout weeks and calculating Weeks of Stock to help optimize inventory policies.
