Printhead maintenance for recommending printhead replacement
Systems and methods of recommending replacement of printheads. In an embodiment, a system trains a first neural network to generate anomaly scores for printheads using an unsupervised learning algorithm based on first training samples of conforming printhead data from a pool of conforming printheads. The system generates a training dataset for a recurrent second neural network by identifying training printhead data for a pool of training printheads, inputting second training samples of the training printhead data into the first neural network to generate training anomaly scores for the training printheads over a plurality of time units, and formatting third training samples for the training printheads. The system trains the recurrent second neural network to generate scaled anomaly scores for printheads using a supervised learning algorithm based on the second training dataset.
1 . A printhead maintenance supervisor, comprising:
at least one processor and memory;
the at least one processor is configured to cause the printhead maintenance supervisor at least to:
train a first neural network to generate anomaly scores for printheads using an unsupervised learning algorithm based on first training samples of conforming printhead data from a pool of conforming printheads;
generate a training dataset for a recurrent second neural network by:
identifying training printhead data for a pool of training printheads;
inputting second training samples of the training printhead data into the first neural network to generate training anomaly scores for the training printheads over a first plurality of time units; and
formatting third training samples for the training printheads, wherein each of the third training samples comprises a first time-series of training data objects over a number of consecutive time units for a training printhead, and a label for the training printhead, and wherein each of the training data objects includes a training anomaly score generated by the first neural network for the training printhead, and at least a subset of the training printhead data for the training printhead; and
train the recurrent second neural network to generate scaled anomaly scores for printheads using a supervised learning algorithm based on the training dataset.
2 . The printhead maintenance supervisor of claim 1 , wherein:
the label represents a printhead condition of a corresponding training printhead.
3 . The printhead maintenance supervisor of claim 1 , wherein:
the subset of the training printhead data in the third training samples comprises a nozzle failure value indicating a number of failed nozzles.
4 . The printhead maintenance supervisor of claim 1 , wherein the at least one processor is configured to further cause the printhead maintenance supervisor at least to:
identify deployed printhead data for a plurality of deployed printheads;
operate the first neural network to generate deployment anomaly scores for the deployed printheads;
operate the recurrent second neural network to scale the deployment anomaly scores generated by the first neural network to produce the scaled anomaly scores for the deployed printheads; and
provide a replacement recommendation for one or more of the deployed printheads based on the scaled anomaly scores for the deployed printheads.
5 . The printhead maintenance supervisor of claim 4 , wherein the at least one processor is configured to further cause the printhead maintenance supervisor at least to:
store the replacement recommendation in a data log corresponding with a deployed printhead.
6 . The printhead maintenance supervisor of claim 4 , wherein the at least one processor is configured to further cause the printhead maintenance supervisor at least to:
transmit a message indicating the replacement recommendation via a network interface.
7 . The printhead maintenance supervisor of claim 4 , wherein the at least one processor is configured to further cause the printhead maintenance supervisor at least to:
display the replacement recommendation via a user interface.
8 . The printhead maintenance supervisor of claim 4 , wherein the at least one processor is configured to further cause the printhead maintenance supervisor at least to:
perform a comparison of one or more of the scaled anomaly scores for each of the deployed printheads to a threshold value;
determine whether the one or more of the deployed printheads are experiencing a failure condition based on the comparison of the scaled anomaly scores; and
provide the replacement recommendation for the one or more of the deployed printheads when the failure condition is determined.
9 . The printhead maintenance supervisor of claim 4 , wherein to operate the first neural network, the at least one processor is configured to further cause the printhead maintenance supervisor at least to:
input first input samples of the deployed printhead data into the first neural network to generate the deployment anomaly scores for the deployed printheads over a second plurality of time units.
10 . The printhead maintenance supervisor of claim 9 , wherein to operate the recurrent second neural network, the at least one processor is configured to further cause the printhead maintenance supervisor at least to:
format second input samples for the deployed printheads, wherein each of the second input samples comprises a second time-series of deployment data objects over the number of consecutive time units for a deployed printhead, and wherein each of the deployment data objects includes a deployment anomaly score generated by the first neural network for the deployed printhead, and at least a subset of the deployed printhead data for the deployed printhead; and
input the second input samples for the deployed printheads into the recurrent second neural network to output the scaled anomaly scores for the deployed printheads.
11 . The printhead maintenance supervisor of claim 1 , wherein:
the first neural network comprises an autoencoder; and
the recurrent second neural network comprises a Long Short-Term Memory (LSTM) neural network.
12 . A cloud computing platform comprising the printhead maintenance supervisor of claim 1 .
13 . A method of recommending replacement of printheads, the method comprising:
training a first neural network to generate anomaly scores for printheads using an unsupervised learning algorithm based on first training samples of conforming printhead data from a pool of conforming printheads;
generating a training dataset for a recurrent second neural network by:
identifying training printhead data for a pool of training printheads;
inputting second training samples of the training printhead data into the first neural network to generate training anomaly scores for the training printheads over a first plurality of time units; and
formatting third training samples for the training printheads, wherein each of the third training samples comprises a first time-series of training data objects over a number of consecutive time units for a training printhead, and a label for the training printhead, and wherein each of the training data objects includes a training anomaly score generated by the first neural network for the training printhead, and at least a subset of the training printhead data for the training printhead; and
training the recurrent second neural network to generate scaled anomaly scores for printheads using a supervised learning algorithm based on the training dataset.
14 . The method of claim 13 , wherein:
the label represents a printhead condition of a corresponding training printhead.
15 . The method of claim 13 , wherein:
the subset of the training printhead data in the third training samples comprises a nozzle failure value indicating a number of failed nozzles.
16 . The method of claim 13 , further comprising:
identifying deployed printhead data for a plurality of deployed printheads;
operating the first neural network to generate deployment anomaly scores for the deployed printheads;
operating the recurrent second neural network to scale the deployment anomaly scores generated by the first neural network to produce the scaled anomaly scores for the deployed printheads; and
providing a replacement recommendation for one or more of the deployed printheads based on the scaled anomaly scores for the deployed printheads.
17 . The method of claim 16 , further comprising:
performing a comparison of one or more of the scaled anomaly scores for each of the deployed printheads to a threshold value; and
determining whether the one or more of the deployed printheads are experiencing a failure condition based on the comparison of the scaled anomaly scores;
wherein providing the replacement recommendation comprises providing the replacement recommendation for the one or more of the deployed printheads when the failure condition is determined.
18 . The method of claim 16 , wherein operating the first neural network comprises:
inputting first input samples of the deployed printhead data into the first neural network to generate the deployment anomaly scores for the deployed printheads over a second plurality of time units.
19 . The method of claim 18 , wherein operating the recurrent second neural network comprises:
formatting second input samples for the deployed printheads, wherein each of the second input samples comprises a second time-series of deployment data objects over the number of consecutive time units for a deployed printhead, and wherein each of the deployment data objects includes a deployment anomaly score generated by the first neural network for the deployed printhead, and at least a subset of the deployed printhead data for the deployed printhead; and
inputting the second input samples for the deployed printheads into the recurrent second neural network to output the scaled anomaly scores for the deployed printheads.
20 . A non-transitory computer readable medium embodying programmed instructions executed by a processor, wherein the instructions direct the processor to implement a method of recommending replacement of printheads, the method comprising:
training a first neural network to generate anomaly scores for printheads using an unsupervised learning algorithm based on first training samples of conforming printhead data from a pool of conforming printheads;
generating a training dataset for a recurrent second neural network by:
identifying training printhead data for a pool of training printheads;
inputting second training samples of the training printhead data into the first neural network to generate training anomaly scores for the training printheads over a first plurality of time units; and
formatting third training samples for the training printheads, wherein each of the third training samples comprises a first time-series of training data objects over a number of consecutive time units for a training printhead, and a label for the training printhead, and wherein each of the training data objects includes a training anomaly score generated by the first neural network for the training printhead, and at least a subset of the training printhead data for the training printhead; and
training the recurrent second neural network to generate scaled anomaly scores for printheads using a supervised learning algorithm based on the training dataset.