Event time characterization and prediction in multivariate event sequence domains to support improved process reliability
A computer implemented method of administering a complex system includes receiving multivariate data from a plurality of sensors of the system in an ambient state. Event sequences in the received multivariate data are identified. The multivariate event sequences are projected to a lower stochastic latent embedding. A temporal structure of the sequences is learned in a lower latent space. A probabilistic prediction in the lower latent space is provided. The probabilistic prediction in the lower stochastic latent space is decoded to an event prediction in the ambient state.
1. A computing device comprising:
a processor;
a network interface coupled to the processor to enable communication over a network;
a storage device coupled to the processor;
an engine stored in the storage device, wherein an execution of the engine by the processor configures the computing device to perform acts comprising:
receiving multivariate data from a plurality of sensors of a system in an ambient state;
identifying event sequences in the received multivariate data;
projecting the multivariate event sequences to a lower stochastic latent embedding;
learning a temporal structure of the sequences in a lower latent embedding;
providing a probabilistic prediction in the lower latent space; and
decoding the probabilistic prediction in the lower stochastic latent space to an event prediction in the ambient state; and
adjusting one or more components of the system to prevent the predicted event.
2. The computing device of claim 1 , wherein the multivariate data comprises event sequences of different variables.
3. The computing device of claim 2 , wherein at least some of the different variables are interrelated.
4. The computing device of claim 2 , wherein the execution of the engine further configures the computing device to perform an act comprising compressing the event sequences into a compact representation of latent variables.
5. The computing device of claim 4 , wherein the execution of the engine further configures the computing device to perform an act comprising using a recurrent neural network (RNN) to capture temporal dynamics of an event history from the compact representation.
6. The computing device of claim 4 , wherein the execution of the engine further configures the computing device to perform an act comprising using learned dynamics in the compact representation to detect a period of anomalous behavior of the system.
7. The computing device of claim 1 , wherein projecting the multivariate event sequences to a lower stochastic latent embedding is by way of a variational auto encoder (VAE).
8. The computing device of claim 1 , wherein the execution of the engine further configures the computing device to perform an act comprising providing an uncertainty of the prediction in the ambient state.
9. The computing device of claim 1 , wherein the system is an industrial plant.
10. The computing device of claim 1 , wherein the variables comprise two or more of pressure, temperature, motion, volumetric flow, or sound level.
11. The computing device of claim 1 , wherein the execution of the engine further configures the computing device to perform an act comprising providing multiple time step prediction by autoregressively predicting latent variables at different future points in time.
12. A non-transitory computer readable storage medium tangibly embodying a computer readable program code having computer readable instructions that, when executed, causes a computer device to carry out a method of administering a system, the method comprising:
receiving multivariate data from a plurality of sensors of the system in an ambient state;
identifying event sequences in the received multivariate data;
projecting the multivariate event sequences to a lower stochastic latent embedding;
learning a temporal structure of the sequences in a lower latent embedding;
providing a probabilistic prediction in the lower latent space;
decoding the probabilistic prediction in the lower stochastic latent space to an event prediction in the ambient state; and
adjusting one or more components of the system to prevent the predicted event.
13. The non-transitory computer readable storage medium of claim 12 , wherein the multivariate data comprises event sequences of different variables.
14. The non-transitory computer readable storage medium of claim 13 , further comprising compressing the event sequences into a compact representation of latent variables.
15. The non-transitory computer readable storage medium of claim 12 , further comprising projecting the multivariate event sequences to a lower stochastic latent embedding is by way of a variational auto encoder (VAE).
16. The non-transitory computer readable storage medium of claim 12 , further comprising providing an uncertainty of the prediction in the ambient state.
17. The non-transitory computer readable storage medium of claim 12 , further comprising providing multiple time step prediction by autoregressively predicting latent variables at different future points in time.
18. A computer implemented method, comprising:
receiving multivariate data from a plurality of sensors of the system in an ambient state;
identifying event sequences in the received multivariate data;
projecting the multivariate event sequences to a lower stochastic latent embedding;
learning a temporal structure of the sequences in a lower latent embedding;
providing a probabilistic prediction in the lower latent space;
decoding the probabilistic prediction in the lower stochastic latent space to an event prediction in the ambient state; and
adjusting one or more components of the system to prevent the predicted event.