NXW2.1P09 - Periodic Suggestions for Port Continuous Improvements
It is a tool centered on algorithms for continuous improvement, generating regular reports that may (or not) be complemented with LLMs explanations. The algorithms are innovative and allow to find good practices and bad practices. It also have indices that reflect how relevant is the need for an improvement. Special care was taken on datasets over terminal operations but other port data can be used.
-
Main Gain(s): Faster identification of operational inefficiencies, Better prioritisation of improvement actions
-
Core Contributions: (concept) E. Rocha; (implementation) E. Rocha; (integration) E. Rocha
Use Cases:
-
Historical data from the several terminals of the Port of Sines
-
Start TRL: 2 - started from scratch.
-
Final TRL: 7 - Validated and tested but no feedback from the Port Authority about its usability and pertinence.
-
Main Features
-
Security and role enforced: Data access through a connection to our Data Warehouse with controlled permissions via Keycloak.
-
Automated continuous-improvement reports: Generates periodic reports with improvement opportunities for port and terminal operations, helping managers regularly identify where processes can be enhanced.
-
Detection of good and bad operational practices: Uses innovative algorithms to analyse operational datasets and identify patterns associated with high-performing practices, inefficiencies, bottlenecks, or recurring poor practices.
-
Improvement-priority indices: Provides indices that quantify how relevant or urgent each improvement opportunity is, allowing teams to prioritise actions based on operational impact.
-
LLM-supported explanations and recommendations: Can complement algorithmic results with natural-language explanations, making the findings easier to understand and translate into practical improvement actions, whenever the indices are high (in order to reduce LLMs´ costs).
Images and Videos
| Demonstration (only on the set of reports "Equipment Operations on Quantity by Hour") |