Unplanned downtime at a casting plant fell from 30 to 1.5 hours a month

Casting depends on drive motors, gearboxes, pumps, and blowers that run continuously. When one stops without warning, the mould flow stops with it, and the line cannot wait.

Maintenance used to be reactive. A failure was found when the machine had already stopped, adding up to about 30 hours of unplanned downtime a month. At roughly Rp 20 million an hour, hundreds of millions of rupiah were lost every month before a repair even began.

Sensors went into six areas: Sand Tank 1 and 2, BE, MM, OC, and moulding pressure. All data stays on the plant’s local network and never goes out to the internet.

Each sensor is compared with its own machine’s baseline and with the ISO 10816 zone limits. When a bearing’s vibration starts to climb, the maintenance team looks at the trend, confirms it against other parameters on the same machine, and schedules the repair.

Spectrum analysis catches early bearing raceway damage 4 to 6 weeks before the vibration can be felt or the bearing runs hot. That is enough time to get the parts ready and pick an off shift, without cutting into the casting schedule.

Over 12 months, unplanned downtime fell from about 30 to 1.5 hours a month. Six potential major failures were handled as scheduled repairs on the off shift instead of as emergencies.

The system does not guarantee that a machine never fails. What changed is the timing: the signs show up earlier, so the team decides when a machine stops, not the machine.