Digital Twin and infrastructural monitoring

Digital Twin for Predictive Maintenance through Image Recognition

An innovative Object and Anomaly Detection system paves the way for the Predictive Maintenance of the national electricity provider. The creation of a Digital Twin of the territory via techniques of 3D modelling opens the stereotypical cartography to a strategic monitoring of infrastructures. The 3D IOT Modelling is operated through Image Recognition technologies to identify anomalies. Thanks to a preliminary classification of buildings, objects and vegetation using artificial intelligence, a virtual inspection of the environment is now possible at any time.

3D modeling of the territory and classification of objects using AI for identify anomalies

Name of project:

3D Power Line Mapping

Duration:

24 months

Year:

2018-2019

Target markets:

3D modeling, IoT

OUR SOLUTION

By law, low- and medium-voltage line operators have to inspect their electrical infrastructures every year in order to identify anomalies that may cause malfunction or inefficiencies. Up until now, human presence was essential to monitor of the entire line. Today, it is possible to reduce the costs connected to physical inspections by realising a virtual model for predictive maintenance. This model is able to:

– Create an automated Anomaly Detection system using the LiDAR scans and photogrammetry

– Manage large amounts of LiDAR laser scans to map the entire infrastructure

– Build a 3D projection through photogrammetry, starting from high-resolution images

– Develop automated extraction processes relating to structural and environmental components through the identification and the gathering of points that are located in the digital space (point cloud)

– Create a web application for digital inspections of the anomalies identified by the software.

Thanks to the point cloud deriving from the LiDAR scans and the photogrammetric reconstruction, Machine Learning algorithms have been implemented for the processing and analysis of these points in the space. The use of artificial neural networks together with Deep Learning indexing models has led to a final point cloud output, where each point belongs to a specific class referring to singular infrastructural components.

Advantages

To satisfy the functional requisites demanded by the Predictive Maintenance of the electricity line, the system developed by Spindox Labs relies on three software components:

Displayer: An application to carry digital inspections, verify anomalies and display the modifications operated by the client.

Repository Management: An application that can upload, manage and index large data relating to mapping.

Process Management: A software of the automated management of the information extracted from images and scans.

The growing performances in terms of 3D environment classification have thoroughly maximised maintenance operations and their simplification. Moreover, the multilingual and interconnected applications have made it possible to automate processes that up until now were manual by necessity.

PARTNERS
Fbk e Spindox Labs digital twin
RELATED PROJECTS

SpinRetail, in Trento the supermarket of the future

DEEP LEARNING, IOT sensors, DATA SCIENCE

Asset tracking for the connected car

 

IOT sensors, Asset Monitoring & Predictive Maintenance

Tech Radar, our Web Crawler

Augmented Intelligence, DEEP LEARNING