DEQUANTIZING LOW-RESOLUTION IOT SIGNALS TO PRODUCE HIGH-ACCURACY PROGNOSTIC INDICATORS

• US 20190310617A1
• Filed: 04/06/2018
• Published: 10/10/2019
• Est. Priority Date: 04/06/2018
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Li, Mengying, Gross, Kenny C.
• Granted Patent: US 10,496,084 B2
• Time in Patent Office: Days
• CPC Class Codes: G05B 23/0221 Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G06F 11/008 Reliability or availability analysis
• CPC Class Codes: G06F 11/3058 Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations thermal management in cooling arrangements of a computing system G06F1/206
• CPC Class Codes: G06F 17/142 Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]

Abstract

The disclosed embodiments relate to a system that removes quantization effects from a set of time-series signals to produce highly accurate approximations of a set of original unquantized signals. During operation, for each time-series signal in the set of time-series signals, the system determines a number of quantization levels (N_QL) in the time-series signal. Next, the system performs a fast Fourier transform (FFT) on the time-series signal to produce a set of Fourier modes for the time-series signal. The system then determines an optimal number of Fourier modes (N_mode) to reconstruct the time-series signal based on the determined N_QL for the time-series signal. Next, the system selects N_mode largest-amplitude Fourier modes from the set of Fourier modes for the time-series signal. The system then performs an inverse FFT operation using the N_mode largest-amplitude Fourier modes to produce a dequantized time-series signal to be used in place of the time-series signal.

Claims

1. A method for removing quantization effects from a set of time-series signals to produce highly accurate approximations of a set of original unquantized signals, comprising:

receiving the set of time-series signals from a monitored system, wherein each signal comprises a sequence of quantized values that were sampled from an original signal generated by the monitored system;

for each time-series signal in the set of time-series signals,

determining a number of quantization levels (N_QL) in the time-series signal;

performing a fast Fourier transform (FFT) on the time-series signal to produce a set of Fourier modes for the time-series signal;

determining an optimal number of Fourier modes (N_mode) to reconstruct the time-series signal based on the determined N_QL for the time-series signal;

selecting N_mode largest-amplitude Fourier modes from the set of Fourier modes for the time-series signal; and

performing an inverse FFT operation using the N_mode largest-amplitude Fourier modes to produce a dequantized time-series signal to be used in place of the time-series signal;

analyzing the set of time-series signals using a prognostic-surveillance system to detect incipient anomalies that arise during execution of the monitored system; and

when the prognostic-surveillance system detects an anomaly, generating an alarm.

2. The method of claim 1, wherein when N_QL is greater than or equal to an acceptable number of quantization levels, the method does not perform a dequantization operation on the time-series signal because the time-series signal already possesses the acceptable number of quantization levels.

3. The method of claim 1, wherein when N_QL is greater than or equal to a minimum fraction of the number of quantized values in the time-series signal, the method does not perform a dequantization operation on the time-series signal because the time-series signal possesses an insufficient number of quantized values to perform the dequantization operation.

4. The method of claim 1, wherein when N_QL=2, instead of performing the FFT and the inverse FFT operations to dequantize the signal, the method:

computes a frequency F_lower for a lower quantization level of the two quantization levels in a moving window that slides across the signal; and

computes the dequantized signal to be 2×

F_lower−

1 assuming the two quantization levels are −

1 and 1.

5. The method of claim 1, wherein determining the number of quantization levels in the time-series signal includes:

sorting the quantized values in the time-series signal to produce a sorted signal;

computing a first-order numerical derivative of the sorted signal; and

determining N_QL to be a number of non-zero derivatives in the first-order numerical derivative of the sorted signal plus one.

6. The method of claim 1, wherein determining N_mode for the time-series signal involves performing a lookup in a library of pre-determined optimal N_mode values based on the determined N_QL for the time-series signal.

7. The method of claim 1, wherein detecting an incipient anomaly comprises detecting an impending failure of the monitored system.

8. The method of claim 1, wherein analyzing the set of time-series signals using the prognostic-surveillance system comprises:

using an inferential model trained on previously received time-series signals from the monitored system to generate estimated values for the time-series signals based on correlations between the time-series signals;

performing a pairwise differencing operation between actual values and the estimated values for the time-series signals to produce residuals; and

performing a sequential probability ratio test (SPRT) on the residuals to detect incipient anomalies that arise during operation of the monitored system.

9. The method of claim 8, wherein the inferential model is trained using a nonlinear, nonparametric (NLNP) regression technique.

10. The method of claim 9, wherein the NLNP regression technique comprises a Multivariate State Estimation Technique (MSET).

11. A system that removes quantization effects from a set of time-series signals to produce highly accurate approximations of a set of original unquantized signals, comprising:

at least one processor and at least one associated memory;

a dequantization mechanism that executes on the at least one processor, wherein during operation, the dequantization mechanism;

receives the set of time-series signals from a monitored system, wherein each signal comprises a sequence of quantized values that were sampled from an original signal generated by the monitored system;

for each time-series signal in the set of time-series signals,

determines a number of quantization levels (N_QL) in the time-series signal;

performs a fast Fourier transform (FFT) on the time-series signal to produce a set of Fourier modes for the time-series signal;

determines an optimal number of Fourier modes (N_mode) to reconstruct the time-series signal based on the determined N_QL for the time-series signal;

selects N_mode largest-amplitude Fourier modes from the set of Fourier modes for the time-series signal; and

performs an inverse FFT operation using the N_mode largest-amplitude Fourier modes to produce a dequantized time-series signal to be used in place of the time-series signal; and

a prognostic-surveillance mechanism that executes on the at least one processor, wherein during operation, the prognostic-surveillance mechanism;

analyzes the set of time-series signals using a prognostic-surveillance system to detect incipient anomalies that arise during execution of the monitored system; and

when the prognostic-surveillance system detects an anomaly, generates an alarm.

12. The system of claim 11, wherein when N_QL is greater than or equal to an acceptable number of quantization levels, the dequantization mechanism does not perform a dequantization operation on the time-series signal because the time-series signal already possesses the acceptable number of quantization levels.

13. The system of claim 11, wherein when N_QL is greater than or equal to a minimum fraction of the number of quantized values in the time-series signal, the dequantization mechanism does not perform a dequantization operation on the time-series signal because the time-series signal possesses an insufficient number of quantized values to perform the dequantization operation.

14. The system of claim 11, wherein when N_QL=2, instead of performing the FFT and the inverse FFT operations to dequantize the signal, the dequantization mechanism:

computes a frequency F_lower for a lower quantization level of the two quantization levels in a moving window that slides across the signal; and

computes the dequantized signal to be 2×

F_lower−

1 assuming the two quantization levels are −

1 and 1.

15. The system of claim 11, wherein determining the number of quantization levels in the time-series signal includes:

sorting the quantized values in the time-series signal to produce a sorted signal;

computing a first-order numerical derivative of the sorted signal; and

determining N_QL to be a number of non-zero derivatives in the first-order numerical derivative of the sorted signal plus one.

16. The system of claim 11, wherein determining N_mode for the time-series signal involves performing a lookup in a library of pre-determined optimal N_mode values based on the determined N_QL for the time-series signal.

17. The system of claim 11, wherein detecting an incipient anomaly comprises detecting an impending failure of the monitored system.

18. The system of claim 11, wherein while analyzing the set of time-series signals, the prognostic-surveillance mechanism:

uses an inferential model trained on previously received time-series signals from the monitored system to generate estimated values for the time-series signals based on correlations between the time-series signals;

performs a pairwise differencing operation between actual values and the estimated values for the time-series signals to produce residuals; and

performs a sequential probability ratio test (SPRT) on the residuals to detect incipient anomalies that arise during operation of the monitored system.

19. The system of claim 18, wherein the inferential model is trained using a Multivariate State Estimation Technique (MSET).

20. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for removing quantization effects from a set of time-series signals to produce highly accurate approximations of a set of original unquantized signals, the method comprising:

receiving the set of time-series signals from a monitored system, wherein each signal comprises a sequence of quantized values that were sampled from an original signal generated by the monitored system;

for each time-series signal in the set of time-series signals,

determining a number of quantization levels (N_QL) in the time-series signal;

performing a fast Fourier transform (FFT) on the time-series signal to produce a set of Fourier modes for the time-series signal;

determining an optimal number of Fourier modes (N_mode) to reconstruct the time-series signal based on the determined N_QL for the time-series signal;

selecting N_mode largest-amplitude Fourier modes from the set of Fourier modes for the time-series signal; and

performing an inverse FFT operation using the N_mode largest-amplitude Fourier modes to produce a dequantized time-series signal to be used in place of the time-series signal;

analyzing the set of time-series signals using a prognostic-surveillance system to detect incipient anomalies that arise during execution of the monitored system; and

when the prognostic-surveillance system detects an anomaly, generating an alarm.