They create a bank of machinery sounds to prevent breakdowns through artificial intelligence

    VALENCIA, 29 Abr. (EUROPA PRESS) -

The ITI - a private technological center specialized in ICT - is working on the 'Soroll-IA2' project, which aims to create an audio database to obtain relevant information that can be used as a predictive maintenance tool and help Prevent possible breakdowns through sound recognition.

Financed by the Valencian Institute of Business Competitiveness (IVACE i) and the European Regional Development Fund (ERDF), the initiative has the participation of the Port of Valencia, Fermax, DCM Automatizada and CASVA Simetría.

One of the main causes of industrial machinery breakdowns, as explained by ITI in a statement, is ignoring warning signs, which can lead to disastrous consequences for any company. Along these lines, a failure of any machine involved in the daily work of a business translates into significant economic losses, but also into delays in production or even possible risks to the safety of workers.

The project, which is in its second year, has been divided into two parts. On the one hand, the recording and capture of audio data in industrial environments and, on the other, the labeling and classification of these. The data is collected through a series of acoustic nodes located in strategic locations and, using Machine Listening techniques, valuable information is extracted. It is an Artificial Intelligence (AI) technology that works and designs algorithms and models capable of extracting knowledge and relevant information from audio signals.

Until now, the main problem was the lack of audio to train the AI ​​models. Therefore, the first phase of the project has focused on collecting the necessary data to create a database of sounds contextualized in the Valencian Community within an industrial environment through a network of Internet of Things (IoT) acoustic sensors.

In this sense, the main characteristic of 'Noise-IA2' is the capture of sounds for subsequent classification in three different types of environments: exteriors, interiors and in the room.

For each type of recording, ITI is collaborating with companies in the industrial sector with the aim of obtaining valuable data that can provide relevant information on a daily basis. The Port of Valencia is providing outdoor sounds framed in a port environment. The companies DCM Automatizada and CASVA Simetry offer the opportunity to record industrial sounds in indoor environments and, with the help of Fermax and its video intercoms, the collection of sounds recorded in the room is being carried out.

Audio data constitutes the information with which AI models are trained so that they perform detection, classification and sound pattern recognition tasks. A possible application of what is obtained thanks to sounds is the identification of sources of noise pollution, but, as Andrea Ceba, manager of the Soroll-IA2 project at ITI, highlights, "what is truly important about the project is the information obtained" .

The sounds, properly processed, provide valuable information that can be used in various areas, although the project is focusing on the use of this data in industrial environments. It should be noted that predictive maintenance is the future objective that we want to achieve thanks to the analysis of sounds that later lead to breakdowns.

"One of the main characteristics of industrial machines is that they make noise, an analysis of these sounds using AI can give us information about the state of the machine and observe if a deviation is occurring with respect to its normal operation. It is expected that predictive maintenance ends up replacing classic preventive maintenance, due to the cost reduction that its use would entail," says, for his part, the principal researcher of the Computer Hearing Research Group at ITI, Pedro Zuccarello.

Within the framework of the project, ITI has created its own recording room in which it is collecting, for example, sounds from Fermax video intercoms with the specific objective of collecting data to train AI algorithms that are capable of recognizing audible alarms.

This room of approximately 11 square meters, located in the new headquarters of the technology center in the Paterna Technology Park, is semi-anechoic. This means that the direct sound path prevails over the reverberations. This condition, added to the high-precision audio equipment, allows the ITI team to obtain top-quality sound evidence. The recording room also grants greater autonomy to ITI to carry out sound recordings in specific projects and better adapt to the possible needs of collaborating companies that require this type of service.