BNSF: Automatic Train Identification: Multidisciplinary Approach to Improve Safety and Efficiency
BNSF Railway has invested significant capital on various track-side detectors to monitor the condition of critical mechanical parts including wheels and bearings to enable proactive maintenance of assets. Among others, we have 1200 Hot Box Detectors (HBD’s) which measure wheel bearing temperature as trains pass by and generate 25,000 messages daily. Many HBD’s are in locations which lack Automatic Equipment Identification (AEI) devices. This presents the unique challenge of accurately matching HBD measurements with the corresponding train, car, and axle. The Operations Research team at BNSF developed a suite of descriptive, predictive, and prescriptive analytics tools that significantly improved train matching efficiency and accuracy. Compared to the legacy system, the new system has improved train matching rates from 75% to 98% and reduced processing time by 70 seconds. Additionally, these algorithms directly eliminated the need to install AEI devices (approximately $150 million) near HBD’s across our network.
IBM: Analytics to Reduce Costs and Improve Quality in Wastewater Treatment
Wastewater treatment is carried out in a complex set of steps, in which the wastewater is treated by means of complex biological, physical, and chemical processes. Today, plants are often operated in a conservative and inefficient risk-averse mode, without the ability to quantify the risk or truly minimize the costs. An innovative Operational control applying descriptive (historical data analysis for a simulation model design and plant state estimation); predictive (wastewater process behavior modeled by a transition probability matrix), and prescriptive analytics (Markov Decision Process) was developed. The system was deployed in Lleida (Spain). Use of the system resulted in a dramatic 13.5 percent general reduction in the plant’s electricity consumption; a 14 percent reduction in the amount of chemicals needed to remove phosphorus from the water; and a 17 percent reduction in sludge production.
IBM: MPA Safer System
With the expected increase in vessel traffic and port capacity, the Singapore Maritime and Port Authority (MPA) has been working to ensure that the future Port of Singapore is safe, secure, efficient and sustainable. Project SAFER, “Sense-making Analytics For maritime Event Response”, is an important component in this effort. A collaboration between MPA and IBM Research, Project SAFER aims to design and develop new analytics capabilities for dramatically increasing the efficiency of maritime operations. The system uses novel cognitive-based analytics leveraging machine learning and entity resolution to provide full situational awareness capability, accurate prediction and intelligence for improving maritime decision-making. Using the SAFER machine-learning-based analytics and vessel prediction models, abnormal and suspicious behaviour is instantly discovered. Based on the extent to which the observed activity of individual or multiple interacting vessels deviates from the modelled behaviour, the event is instantly geo-localized, and sent in the form of an alert. MPA can thus address infringements across all 1000 vessels in real-time SAFER system’s automated movement detection leads to a significant accuracy improvement of 34%. Vessel movement information is needed not only for ensuring safety and security but for many other functions including billing: the accuracy improvement achieved by the SAFER system thus has direct implications on revenue and reducing disputes.
Macys: A Model Driven Approach to Store Selling Space Optimization
Store Locations sales performance by merchandise business was until now being compared against benchmarks formulated by averages of similar scale. A new approach has been developed integrating Exploratory Analytics (Co-clustering) and Prescriptive Analytics (Non‐Linear Spline Regression Optimization Model and Seasonal (Random Walk) Autoregressive Integrated Moving Average (SARIMA) Model). We have developed a new workflow to integrate all three models in recommending optimal store layouts and merchandize mix for new store locations and major remodels of existing ones. In this three-tiered process, the analyst first identifies the statistical cluster membership of the under analysis location and formulates a plan based on that benchmark. Then he/she invokes the optimization model that provides the space adjustment recommendations that maximize its sales potential based on existing cross‐sectional data (for remodel stores). In the final step, the forecasting model is used to validate whether the recommendations made based on cross‐sectional (historical) data hold true in the time‐frame where these changes (projected store opening or remodel completion) are expected to take place.
Northwestern University: SAFE (Situational Awareness for Events): A Data Visualization System
Marathons and other endurance events are growing in popularity, and thus require significant resources to ensure safety and success. Event management tools have not grown to meet this need. A team of Northwestern University faculty and students and staff members of the Bank of America Chicago Marathon has developed a data visualization system that incorporates critical data into a user-friendly dashboard to provide a centralized source of information at mass gathering events. This system uses descriptive, predictive and prescriptive analytics to help race organizers and relevant stakeholders effectively manage and oversee all participants, monitor the dynamic location of race participants, and manage health and safety resources throughout the event should any emergency issues arise. Our system is the first comprehensive dashboard for endurance event management. The system provides a dynamic representation of the flow of people and resources. The system integrates real-time dynamic data from tracking devices and predictive algorithms developed by the research team, and presents the information on a summary visual device, both as a large screen in an incident command facility for group monitoring and a desktop/mobile version for individual monitoring. The system has become an integral component in the management of the Bank of America Chicago Marathon and Shamrock Shuffle 8K and the Chevron Houston Marathon and Aramco Half Marathon.
Schneider: Chassis Leasing and Selection Policy for Port Operations
Port cargo drayage operations manage the movement of shipping containers that arrive and depart on ocean-going container vessels and are transported over the road to and from inland trans-loading facilities. While containers are on land they are placed on wheeled chassis until they return to the port facility. A significant operational challenge is the acquisition and management of these chassis. While many port drayage operators simply lease chassis on a per day basis as demand warrants, Schneider National has determined that an analytics-driven policy that combines long term leasing with daily rental leads to significant cost savings while improving both service and reliability. We present and implement a solution methodology that addresses the two decision problems that arise with this dual sourcing approach: 1) the optimal fleet size for leased chassis and 2) a real-time decision policy for selecting between rental and leased chassis as containers are received. As we demonstrate our solution represents an integrated approach that combines the three general areas of analytics methodology and incorporates a particularly novel interplay of optimization, simulation, and predictive modeling. We conclude with an analysis of the financial benefit that has been achieved and a discussion of the applicability of our methodology to other problem settings.