The invention provides an equipment health detection method and device based on joint learning. The method comprises the following steps: responding to device data of a target device with a detection demand sent by a participant in the joint
learning architecture; extracting key
feature data in the equipment data sent by the participants according to the detection requirements of the participants; compressing key
feature data in the equipment data according to the attribute information of the target equipment; establishing an equipment health trend curve by using key
feature data in the compressed equipment data; matching the equipment data of the target equipment with values on the equipment health trend curve; and determining the health state of the predicted target equipment according to a matching result of the equipment data of the target equipment and the values on the trend curve. According to the method and the device, the problems of operation risk and resource waste caused by incapability of timely and accurately detecting and maintaining the equipment in the prior art are solved.