Methods for improved forewarning of critical events across multiple data channels

• US 7,209,861 B2
• Filed: 09/22/2003
• Issued: 04/24/2007
• Est. Priority Date: 07/12/2002
• Status: Expired due to Fees

• Current Assignee: UT-Battelle LLC
• Original Assignee: UT-Battelle LLC
• Inventors: Hively, Lee M.
• Primary Examiner(s): Barbee; Manuel L.
• Application Number:
  US10/667,045
• Publication Number:
  US 20040059761A1
• Time in Patent Office: 1,310 Days
• Field of Search: 702/183, 702/92, 702/22, 702/182, 600/300, 600/544, 600/545
• US Class Current: 702/183
• CPC Class Codes: A61B 5/4094 Diagnosing or monitoring seizure diseases, e.g. epilepsy
• CPC Class Codes: A61B 5/412 Detecting or monitoring sepsis
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/7275 Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06F 2218/02 Preprocessing
• CPC Class Codes: G06F 2218/12 Classification; Matching
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders

Abstract

This disclosed invention concerns improvements in forewarning of critical events via phase-space dissimilarity analysis of data from mechanical devices, electrical devices, biomedical data, and other physical processes. First, a single channel of process-indicative data is selected that can be used in place of multiple data channels without sacrificing consistent forewarning of critical events. Second, the method discards data of inadequate quality via statistical analysis of the raw data, because the analysis of poor quality data always yields inferior results. Third, two separate filtering operations are used in sequence to remove both high-frequency and low-frequency artifacts using a zero-phase quadratic filter. Fourth, the method constructs phase-space dissimilarity measures (PSDM) by combining of multi-channel time-serial data into a multi-channel time-delay phase-space reconstruction. Fifth, the method uses a composite measure of dissimilarity (C_i) to provide a forewarning of failure and an indicator of failure onset.

Claims

1. A method for processing data to provide a forewarning of a critical event, comprising:

acquiring a plurality of sets of data for a plurality of channels of data from monitoring at least one physical test subject or physical process;

calculating a set of channel data for a selected parameter from the plurality of channels of data representing parameters that are calculated to provide the selected parameter;

computing a renormalized measure of dissimilarity from distribution functions derived from a phase space for a selected channel of data;

comparing said renormalized measure of dissimilarity to a threshold (U_C) for a number of occurrences (N_OCC) to indicate a condition change in said renormalized measure of dissimilarity; and

detecting a simultaneous condition change in a plurality (N_SIM) of renormalized measures of dissimilarity to determine a forewarning of the critical event; and

providing an output of at least one of a graph, a table of data or an observable signal by which a human observer can detect the forewarning of the critical event.

2. The method of claim 1, wherein the test subject is a human patient.

3. The method of claim 1, wherein the test subject is a mechanical device or physical process.

4. The method of claim 1, wherein the selected parameter is three-phase electrical power.

5. The method of claim 1, wherein the selected parameter is vibration mechanical power.

6. The method of claim 1, wherein the selected parameter is a difference between two channels of EEG data.

7. The method of claim 1, further comprising:

performing a first filtering of each set of data with a zero-phase quadratic filter that filters out high-frequency artifacts; and

performing a second filtering of each set of data with a zero-phase quadratic filter to filter out low-frequency artifacts.

8. The method of claim 1, further comprising:

sorting data values into ascending order from a minimum to a maximum;

determining a number of unique data values (n) and a corresponding relative occurrence frequency (F_k) for each unique data value (v_k);

displaying a graph of frequency (F_k) versus data values (v_k) in each bin in a connected phase space; and

discarding data that has [v_k−

(N/n)]/σ

₃>

z, where the value of z is determined by solving 1/n=½

erfc(z/>

2), and σ

₃ is a standard deviation in the occurrence frequency.

9. The method of claim 1, with an alternative embodiment for event forewarning, comprising determining a sequence of renormalized phase space dissimilarity measures from data sets for the test subject or process;

summing said renormalized measures into a composite measure, C_i, for an i-th data set;

performing a least-squares analysis over a window of m points of the said composite measure to obtain a straight line, y_i=ai+b, that best fits said composite data in a least-squares sense;

determining a variance, σ

₁²=Σ

_i (y_i−

C_i)²/(m−

1), of said composite measure with respect to the straight line fit;

obtaining a variability of the sequel window of m sequential points via the statistic, G=Σ

_i (y_i−

C_i)²/Σ

₁²;

comparing said value of G to a running maximum value of the same statistic, G_max;

determining the forewarning of or failure onset of a critical event (such as a machine failure), when G is significantly more than G_max;

obtaining a ratio, R=(G_max)_k/(G_max)_k-1, of the present and previous running maximum in G; and

determining the forewarning of a critical event when R is significantly more than some limit.

10. A method for processing data to provide a forewarning of a critical event, comprising:

acquiring a plurality of sets of data for at least two channels of data from monitoring at least one physical test subject or physical process;

producing a set of multi-channel data representing a combination of said at least two channels of data;

computing a multi-channel time-delay phase-space (PS) construction, which has a form;

y(i)=[s(1)₁, s(1)_1+λ

, s(1)_i+2λ

, . . . , s(2), s(2)_i+λ

, s(2)_i+2λ

, . . . , s(c)_i, s(c)_i+λ

, s(c)_i+2λ

, . . . ], where s(c) denotes symbolized data for a c-th channel;

computing a renormalized measure of dissimilarity from distribution functions derived from a phase space for the multi-channel data;

comparing said renormalized measure of dissimilarity to a threshold (U_C) for a number of occurrences (N_OCC) to indicate a condition change in said renormalized measure of dissimilarity; and

detecting a simultaneous condition change in a plurality (N_SIM) of renormalized measures of dissimilarity to determine a forewarning of the critical event; and

providing an output of at least one of a graph, a table of data or an observable signal by which a human observer can detect the forewarning of the critical event.

11. The method of claim 10, further comprising:

performing a first filtering of each set of data with a zero-phase quadratic filter that filters out high-frequency artifacts; and

performing a second filtering of each set of data with a zero-phase quadratic filter to filter out low-frequency artifacts.

12. The method of claim 10, using an alternative embodiment for event forewarning, comprising determining a sequence of renormalized phase space dissimilarity measures from data sets collected during increasingly severe fault conditions;

summing said renormalized measures into a composite measure, C₁, for an i-th data set;

performing a least-squares analysis over a window of m points of the said composite measure to obtain a straight line, y_i=ai+b, that best fits said composite data in a least-squares sense;

determining a variance, σ

₁²=Σ

_i(y_i−

C_i)²/(m−

1), of said composite measure with respect to the straight line fit;

obtaining a variability of a sequel window of m sequential points via the statistic, G=Σ

_i(y_i−

C_i)²/σ

₁²;

comparing said value of G to a running maximum value of the same statistic, G_max; and

determining an onset of a critical event, such as forewarning of a machine failure, when G is significantly more than G(non-end-of-life), or when R is significantly more than R(non-end-of-life), or detection of failure onset when G is significantly greater than G(end-of-life).

13. The method of claim 10, wherein the test subject is a human patient.

14. The method of claim 10, wherein the test subject is a mechanical device or physical process.

15. The method of claim 10, further comprising:

sorting the data values into ascending order from a minimum to a maximum;

determining a number of unique data values (n) and a corresponding relative occurrence frequency (F_k) for each unique data value (v_k);

displaying a graph of frequency (F_k) versus data values (v_k) in each bin in a connected phase space; and

discarding data that has [v_k−

(N/n)]/σ

₃>

z, where the value of z is determined by solving 1/n =½

erfc(z/√

2), and σ

₃ is the standard deviation in the occurrence frequency.

16. The method of claim 10, wherein the multi-channel time-delay phase-space (PS) construction is constructed from process-indicative data, which is three-phase electrical power data.

17. The method of claim 10, wherein the multi-channel time-delay phase-space (PS) construction is constructed from process-indicative data, which is vibration mechanical power data.

18. The method of claim 10, wherein the multi-channel time-delay phase-space (PS) construction is constructed from process-indicative data representing a difference between the two channels of data, which is EEG data.