Objective is to create a marketplace of Artificial Intelligence models for detecting failures and anomalies
The launch of a new car is preceded by a series of road tests, with simulations of real situations, aiming to guarantee the quality of the product and the reduction of flaws that can force the consumer to resort to the manufacturer's guarantee - or even lead to the recall of vehicles. With the aim of improving this process, Previsiown, a startup that operates in the area of road tests for the automotive industry, is starting a new project in partnership with CPQD – and the support of EMBRAPII – aiming to add intelligence to the analysis of collected data. during these tests.
Datalogger prototype to be integrated into AI data analysis proof of concept
This is an evolution of the work being carried out by the partnership Previsiown and CPQD, focusing on the development of a solution for acquiring data from motor vehicles (telemetry) during road tests. The solution consists of on-board diagnostic devices that collect data from the cars (directly from the CAN bus) and, through smartphones, send the information for processing on the IoT platform developed by CPQD - dojot -, which makes them available on the Previsiown backoffice.
Datalogger prototype to be integrated into AI data analysis proof of concept
The solution also records audio comments from drivers carrying out vehicle tests. “During the test, the driver narrates all relevant events, such as noise, power or malfunction of some equipment, for example, which are delivered already classified”, explains Ivan Vianna, CEO of Previsiown.
In addition, in previous projects, the solution incorporated prediction models and pattern recognition, using Artificial Intelligence (AI) resources, based on data collected during the running tests – and made available through dojot. “As a result, we gained speed in the testing process and reduced rework, which was reflected in savings in resources, especially human resources”, says Vianna. “The solution also brought a tenfold increase in the number of occurrences mapped during the tests, which made it possible to quickly identify and deal with problems, directly reflecting on the quality of the final product”, he adds.
With the new project, which has a duration of 12 months, the objective is to generate and integrate several AI models in a marketplace for detecting faults and anomalies in vehicles. “The intention is to create and previously train the AI models, which will be available for consultation in the marketplace”, emphasizes the CEO of Previsiown. The project also foresees the assembly of 50 prototypes of dataloggers (data collectors), which will be integrated in a proof of concept of data analysis by AI models.
