CROSS-DOMAIN TIME SERIES DATA CONVERSION APPARATUS, METHODS, AND SYSTEMS

• US 20160371316A1
• Filed: 06/19/2015
• Published: 12/22/2016
• Est. Priority Date: 06/19/2015
• Status: Active Grant

• Current Assignee: Preferred Elements Incorporated
• Original Assignee: Preferred Networks Incorporated
• Inventors: Okanohara, Daisuke, Clayton, Justin B.
• Granted Patent: US 10,460,251 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Apparatus, methods, and systems for cross-domain time series data conversion are disclosed. In an example embodiment, a first time series of a first type of data is received and stored. The first time series of the first type of data is encoded as a first distributed representation for the first type of data. The first distributed representation is converted to a second distributed representation for a second type of data which is different from the first type of data. The second distributed representation for the second type of data is decoded as a second time series of the second type of data.

Claims

1. A method comprising:

receiving a first time series of a first type of data;

storing the first time series of the first type of data

encoding the first time series of the first type of data as a first distributed representation for the first type of data;

converting the first distributed representation to a second distributed representation for a second type of data which is different from the first type of data; and

decoding the second distributed representation for the second type of data as a second time series of the second type of data.

2. The method of claim 1, wherein the first time series of the first type of data is infrared thermography and the second time series of the second type of data is ultrasound decibel levels.

3. The method of claim 2, wherein the first time series of the first type of data is provided by an infrared camera monitoring a steam trap.

4. The method of claim 3, further comprising inputting the second time series of the second type of data into a detection model.

5. The method of claim 4, wherein the detection model determines that the steam trap is faulty using the second time series of the second type of data.

6. The method of claim 1, further comprising inputting the second time series of the second type of data into a prediction model.

7. The method of claim 6, further comprising transmitting an output of the prediction model based on the second time series of the second data type to a user device.

8. The method of claim 1, further comprising transmitting at least one of the first distributed representation and the second distributed representation from a first device to a second device.

9. The method of claim 1, further comprising:

encoding the second time series of the second type of data as the second distributed representation for the second type of data;

converting the second distributed representation to a third distributed representation for a third type of data which is different from the first type of data and the second type of data; and

decoding the third distributed representation for the third type of data as a third time series of the third type of data.

10. The method of claim 9, further comprising transmitting at least one of the first distributed representation, the second distributed representation, the second time series, the third distributed representation, and the third time series from a first device to a second device.

11. The method of claim 1, wherein converting the first distributed representation to the second distributed representation includes:

converting the first distributed representation to a third distributed representation for a third type of data which is different from the first type of data and the second type of data; and

converting the third distributed representation to the second distributed representation without decoding the third distributed representation.

12. The method of claim 1, wherein the second type of data is an intermediate data type between the first type of data and a third type of data, which a prediction model is configured to receive as an input, and wherein a plurality of different sequences of data type conversions exist between the first type of data and the third type of data,

further comprising determining a sequence of data type conversions to perform from the plurality of different sequences.

13. The method of claim 12, wherein the sequence of data type conversions includes at least two data type conversions provided in a hub device.

14. The method of claim 1, wherein the first type of data and the second type of data are both in the same domain of data, and the first type of data is collected from a different source than the second type of data.

15. The method of claim 14, wherein the different source is at least one of manufactured by a different manufacturer and positioned differently relative to equipment being monitored.

16. An apparatus comprising:

a data collection device configured to collect a first type of data over a period of time;

a memory configured to store time series data collected by the data collection device;

an encoder configured to convert the first time series of the first type of data into a first distributed representation for the first type of data;

a data type converter configured to convert the first distributed representation into a second distributed representation for a second type of data which is different from the first type of data; and

a decoder configured to convert the second distributed representation for the second type of data into a second time series of the second type of data.

17. The apparatus of claim 16, wherein configuring the encoder includes training the encoder on a first set of training data of the first type of data, configuring the decoder includes training the decoder on a second set of training data of the second type of data, and configuring the data type converter includes training the data type converter on a third set of training data including pairs of distributed representations of the first type of data and the second type of data.

18. The apparatus of claim 16, wherein the encoder includes an input layer and a plurality of encoder layers and the decoder includes an output layer and a plurality of decoder layers.

19. An apparatus comprising:

a memory configured to store time series data collected by a data collection device and distributed representation data;

a first encoder configured to convert a first time series of the first type of data into a first distributed representation for the first type of data;

a first decoder configured to convert the first distributed representation for the first type of data into the first time series of the first type of data;

a first data type converter configured to convert the first distributed representation into a second distributed representation for a second type of data which is different from the first type of data;

a second data type converter configured to convert the second distributed representation into the first distributed representation;

a second encoder configured to convert a second time series of the second type of data into the second distributed representation for the second type of data;

a second decoder configured to convert the second distributed representation for the second type of data into the second time series of the second type of data;

a third data type converter configured to convert the first distributed representation into a third distributed representation for a third type of data which is different from the first type of data and the second type of data;

a fourth data type converter configured to convert the third distributed representation into the first distributed representation;

a third encoder configured to convert a third time series of the third type of data into the third distributed representation for the third type of data; and

a third decoder configured to convert the third distributed representation for the third type of data into the third time series of the third type of data.

20. The apparatus of claim 19, wherein the apparatus is one device in a chain of a plurality of different devices.