DINOS (Diverse INdustrial Operation Sounds) is a large-scale, open-access dataset consisting of over 74,000 audio samples totaling more than 1,093 hours, collected from a wide range of industrial acoustic scenarios. It covers diverse manufacturing processes, materials, and operating conditions to comprehensively represent industrial sound characteristics. The dataset includes recordings from CNC cutting operations, additive manufacturing (AM) processes, and designed anomaly scenarios. For cutting, data were collected from two CNC machines: Haas VF-2 and Yornew VMC-300. The VF-2 recordings capture inactive, machining, and warm-up states, while the VMC-300 machines aluminum (Al-6060) under varying spindle speeds and feed rates to induce chatter—a self-excited vibration that excites the system’s natural frequencies, degrading surface finish and tool life. Additional, unlabeled machining sounds were acquired from an APEC SK2540 CNC system. For AM processes, DINOS includes data from Renishaw’s LPBF and FormAlloy’s DED machines, capturing both idle and operational states, with events like fan activation, axis motion, and laser firing reflected in the acoustic signatures. Cold spray data were recorded using both a stethoscope sensor attached to the powder feeder and a microphone in an open area, enabling detection of faults such as gas flow interruptions, powder clogging, and feedstock depletion. Finally, a microphone installed on a shop floor with multiple machines captured ambient industrial sounds, including tool operations, fan noise, and high-pressure air discharge.
