FLEET MANAGEMENT

PROBLEM STATEMENT

Typically, fleet management operations worldwide face major challenges to run successfully. They have high operating expenses and significant maintenance issues which can be detrimental to the business if not addressed properly in time. Sometimes, timely detection of an upstream problem such as engine failures can prevent more expensive downstream issues. Un-scheduled maintenance issues might disrupt Supply Chain deliveries. In order to address these challenges, companies are turning to Ai to avoid catastrophic failures in the fleet vehicles.
The objective of this project is to predict the failure of their vehicles at an early stage and act accordingly to avoid shipment deliveries. The optimum solution requires real-time monitoring from remote vehicles and datadriven vehicle fleet maintenance software can alert a fleet manager to take appropriate actions required to prevent delivery disruptions.

KEY BENEFITS (MINSKY™)

• User-Friendly, cloud based AI platform
• No coding skills are required for results or predictions.
• Provides you a list of % dependency features that can be used to optimize your business
• Ability to fine tune or optimize the models by trying different algorithms / prediction attributes
• Easy integration with other third party solutions such as TABLEAU for data visualization

BENEFITS

• Reduced downtime and increased vehicle availability
• Increased efficiency of fleet and drivers
• Improved supply chain management and on-time deliveries
• Fuel economy and smart resource planning
• Improved driver safety
• Reduction in Maintenance costs
• Better Management of spare parts inventory
• Systematically schedule the optimal maintenance / inspection routine
• Improved productivity in operations
• Improved vehicle lifespan
• Avoid unnecessary scheduled maintenance

SOLUTION OVERVIEW:

Our solution was to explore how artificial intelligence can lead to improved metrics across all assets/vehicles that require routine maintenance and have a positive impact on our customers, employees and the profits. After thoroughly evaluating the client’s challenges/data, we used Minsky™ to accurately model historical data of the fleet-downtime instances along with past data parameters for the vehicles. This process included the collection of vehicle data using IoT sensors remotely from the moving vehicles such as truck id, vehicle speed sensor, engine load, vibration, speed GPS, trip distance, Trip time journey, longitude and latitude of the truck location and other engine related data. Once the AI models were generated by Minsky™ for the selected algorithms and historical data parameters for the vehicles along with past actual vehicle downtimes, predictions for future maintenance were made using real-time equipment data which is gathered from IoT sensors and uploaded to our secure cloud. Prediction data from Minsky™ was also integrated with 3rd party data visualization applications like Tableau.

This process was done in 3 steps:

Step1: AI Modelling from historical data using Minsky™
Step2: Real-time Data acquisition from remote automobile sensors using IoT
Step3: Use analytics to predict future maintenance requirements.

Our Typical Ai data flow:

Minsky™ Results:

Based on the data from Fig 1 above, our analysis shows that Gradient Boosting Classifier (84.4%) and Random Forest Classifier (84.4%) gave the highest top two model accuracies. In terms of the truck data, Trip time journey showed the highest dependency to predict failures in the trucks followed by Trip Distance and Longitude (location).