Real-time anomaly detection in three dimensional printers
AssigneePeridot Print LLC
InventorsAmit Kumar, Hariharan Rajaram, Debjit Roy, Prakash Reddy
Life of this patent
Estimating the term…
Abstract
Examples of systems for detecting anomaly in a print job performed by a three-dimensional printer are described herein. In an example, a data set pertaining to a set of layers printed based on a print job of the 3D printer may be processed by an anomaly detection engine. Thereafter, a data set of a layer being printed by the 3D printer may be obtained. Based on the data set of the set of layers and the data set of the layer being printed, an anomaly may be detected in real-time.
Claims15
A system comprising:
- a processor communicatively coupled to a three-dimensional (3D) printer;
- and
- an anomaly detection engine, coupled to the processor, to:
- receive a first data set pertaining to a set of previously printed layers based on a previous print job of the 3D printer,
- process the first data set through a recurrent neural network (RNN), wherein the first data set comprises data from sensors associated with the 3D printer and attributes pertaining to each layer from the set of previously printed layers;
- cause the 3D printer to print a current print job;
- obtain, in real-time, a second data set of a layer being printed by the 3D printer in the current print job;
- and
- input and process the second data set through the RNN to detect an anomaly in the layer being printed by the 3D printer.
A method comprising:
- causing, by a processor, a three-dimensional (3D) printer to print a current print job;
- obtaining, by the processor, from the 3D printer, a first data set pertaining to a set of layers printed based on a previous print job of the 3D printer and a second data set pertaining to a layer being printed by the 3D printer in the current print job;
- providing the first data set and the second data set as an input to an encoder-decoder based model, wherein an encoder is to generate a hidden state representation of the first data set and the second data set, and a decoder is to detect an anomaly in the layer being printed in the current print job;
- and
- notifying a user, in real-time, about the detected anomaly.
A non-transitory computer-readable medium comprising computer-readable instructions, which, when executed by a processing resource of a system, cause the processing resource to:
- provide a print command to a three-dimensional (3D) printer to cause the 3D printer to print a current print job;
- receive, a data set pertaining to a set of layers previously printed by the 3D printer in a previous print job, and another data set pertaining to a layer being printed by the 3D printer in the current print job;
- provide the data set to an anomaly detection engine to identify a deviation in the another data set, the anomaly detection engine being trained over an encoder-decoder based recurrent neural network (RNN);
- upon determination that the deviation exceeds above a predefined threshold value, detect an anomaly in the layer being printed by the 3D printer in the current print job;
- and
- generate, in real-time, a notification informing degradation in a layer quality associated with the current print job of the 3D printer.
Dependencies are inferred from the claim text.