Complex technologies enabling Industry 4.0, such as Acoustic Artificial Intelligence, interconnection and computational automation and data analysis operated in the cloud, are beginning to make large-scale industrial processes more efficient.
Specifically AiProd, the innovative application of Acoustic Artificial Intelligence, patented by the Pisa-based IProd, combines, in real time, the multimedia files associated with IoT tags generated by industrial assets or products being tested with the patterns of Artificial Intelligence in in order to enable the creation of clusters capable of training a necessary and sufficient number of neural networks to perform more and more accurate surveys which can be associated with homogeneous processing or testing scenarios.
Thanks to the AiProd application, IProd with its Tornado Team won the "AIRtificial Intelligence" Challenge, the 48-hour marathon dedicated to Artificial Intelligence, organized by the Air Force together with Leonardo at the Institute of Military Aeronautical Sciences of Florence, center of excellence for the formation of the leadership of the Armed Force.
AIRtificial Intelligence is the hackathon carried out in Italy as part of a broad innovation process concerning various organizational and functional fields, in particular regarding the theme of Logistics 4.0
AIRtificial Intelligence represented an unmissable opportunity to harmonize an ecosystem of teams of specialists with a high technological and innovative profile in which start-ups, universities, industry, agencies and public organizations and sector experts, have pooled their knowledge, skills and talent in the field of Artificial Intelligence and Machine Learning.
Nine teams challenged each other in the creation of virtual assistance solutions, using Artificial Intelligence methodologies applicable to aeronautical and aerospace systems. The teams discussed the interaction of AI both with the issue of maintenance training, which thanks to the use of virtual platforms can be significantly optimized and speeded up, and with remote maintenance activities through the support of a virtual assistant.
The Tornado Team won the first prize of the Challenge, thanks to the aim of automating the manual Tapping Test operations on a helicopter blade through the AiProd application based on Acoustic Artificial Intelligence, self-trained, preventing the results on the structural integrity of this component in order to optimize maintenance interventions.
Specifically, the Tornado Team, thanks to the AiProd application, has developed a system that allows the inspector to identify when a helicopter blade is close to failure. This component is made of composite material, with a very sophisticated internal stratification in which the carbon fibers are oriented according to the direction of the stresses generated during operation. The carbon fibers, despite their inherent structural stiffness and their very high mechanical resistance to traction, progressively detach with the progress of the number of flight hours until they cause the blade to break permanently, an event that must obviously be avoided.
To prevent breakage, the helicopter blade is subjected to the so-called "Tapping Test" which is performed with a predefined planning, depending on the number of flight hours. During the test, the inspector acts with a special hammer hitting a predetermined area of the blade surface, listening to the noise generated by this operation and signaling the need for maintenance on the component as a result of an abnormal sound generated during the test.
To automate this process, the AiProd application intervenes, which listens to the sound generated by the Tapping Test and, with the aid of a camera that records the blade during the hammering operation, based on the point where it is hit, trains a neural network that exactly recognizes the sound generated by a new blade, in the absence of structural anomalies, from that generated by a blade characterized by a certain duration of use. In fact, when a helicopter blade is hammered, anomalies are found, point by point, on a scale from 0% to 100%, where 100% identifies the conformity of the structural behavior in the absence of defects, while 0% identifies, compared to a reference situation, a strong anomaly in the structural behavior of the blade at that particular point.
AiProd opens interesting application domains, being able to predict outcomes on the quality of a work in progress or on the quality of a product being tested, helping manufacturers to prevent costly production waste or costly downtime for technical assistance and users. final to optimize the life cycle of use of the products and components used in their processes.
