Résumé
The idea of using the piezoelectric element of an inkjet printhead as an acoustic sensor for inferring the status of the jetting nozzles is almost as old as inkjet itself. While piezoelectric inkjet printing devices have evolved considerably since the early days of inkjet printing, enabling the continuous development of novel inks, functional fluids, substrates, pre- and post- printing treatments, and facilitating the adoption of new material deposition processes across many industries, nozzle acoustic sensing has seen minimal adoption by the industry. With a few notable exceptions, the inkjet community has been reticent to adopt this aspect of the inkjet technology. The main reasons argued for not embracing this technology are its perceived inability to identify subtle failure modes (deviated nozzles) and the difficulties of machine operators to interpret and react to this novel type of information. In this paper we will argue that developments in Artificial Intelligence can help overcome these limitations. Agentic AI and Reinforcement Learning provide a conceptual framework and a technology capable of improving nozzle failure classifiers and defining and evaluating multiple automatic responses of the printing system to changing printing conditions, enabling a quasi-real-time optimization of the printing process.