Method for verifying bad pattern in time series sensing data and apparatus thereof

• US 9,547,544 B2
• Filed: 06/17/2014
• Issued: 01/17/2017
• Est. Priority Date: 06/18/2013
• Status: Active Grant

• Current Assignee: Samsung SDS (China) Co., Ltd. (Samsung Group)
• Original Assignee: Samsung SDS (China) Co., Ltd. (Samsung Group)
• Inventors: Shin, Kae Young, Ahn, Dae Jung, Seo, Dae Hong, Jung, Woo Young
• Primary Examiner(s): Decady, Albert
• Assistant Examiner(s): Alshack, Osman M
• Application Number:
  US14/306,967
• Publication Number:
  US 20140372813A1
• Time in Patent Office: 945 Days
• Field of Search: 714/704, 714/799, 714/705
• US Class Current: 1/1
• CPC Class Codes: G01R 31/3171 BER [Bit Error Rate] test
• CPC Class Codes: G06F 11/0703 Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation
• CPC Class Codes: G06F 11/0706 the processing taking place on a specific hardware platform or in a specific software environment
• CPC Class Codes: G06F 11/0736 in functional embedded systems, i.e. in a data processing system designed as a combination of hardware and software dedicated to performing a certain function testing or monitoring of automated control systems G05B23/02
• CPC Class Codes: G06F 11/076 by exceeding a count or rate limit, e.g. word- or bit count limit
• CPC Class Codes: G06F 11/0772 Means for error signaling, e.g. using interrupts, exception flags, dedicated error registers
• CPC Class Codes: G06F 11/079 Root cause analysis, i.e. error or fault diagnosis in a hardware test environment G06F11/22; in a software test environment G06F11/36
• CPC Class Codes: H03M 13/015 Simulation or testing of codes, e.g. bit error rate [BER] measurements

Abstract

A method for verifying bad pattern in time series sensing data by calculating a bad pattern error rate, which can be applied to the time series sensing data measured and produced from a predetermined sensor provided in predetermined equipment, and an apparatus thereof are provided. The method includes receiving information on the bad pattern applied to time series sensing data measured by a suspicious sensor, accessing the time series sensing data of each product, generated by the suspicious sensor during a verification period, calculating similarity measures between the bad pattern based on the bad pattern information and the time series sensing data for each product, and calculating an error rate of the bad pattern based on the similarity measures.

Claims

1. A method for verifying a bad pattern in sensing data, the method comprising:

receiving bad pattern information applied to sensing data measured by a sensor;

accessing the sensing data of each product, generated by the sensor;

calculating similarity measures between the bad pattern based on the bad pattern information and the accessed sensing data; and

calculating an error rate of the bad pattern based on the similarity measures,

wherein the calculating of the error rate of the bad pattern comprises;

querying information of each product; and

calculating the error rate of the bad pattern by comparing the information with at least one of the calculated similarity measures,

wherein the calculating the similarity measures between the bad pattern based on the bad pattern information and the accessed sensing data comprises calculating a first similarity measure between the bad pattern based on the bad pattern information and the sensing data according to a first standard and a second similarity measure between the bad pattern based on the bad pattern information and the sensing data according to a second standard,

wherein each of the first and second similarity measures have a value in a range between 0, which means non-similarity, and 1, which means sameness, and

wherein the calculating of the bad pattern error rate comprises;

when one or more of the first and second similarity measures of a particular product have a value 1 and the particular product is determined as a good product based on the information, selecting the particular product as an error case of the bad pattern information; and

calculating the error rate based on a number of error cases.

2. The method claim 1, wherein the accessing to the sensing data of each product comprises accessing sensing data of each product as fault detection &

classification (FDC) data generated by the sensor for a predetermined period.

3. The method of claim 1, wherein the sensing data is time series sensing data.

4. The method of claim 1, wherein the accessing of the sensing data of each product, generated by the sensor occurs during a verification period.

5. A method for verifying a bad pattern in sensing data, the method comprising:

receiving bad pattern information applied to sensing data measured by a sensor;

accessing the sensing data of each product, generated by the sensor;

calculating similarity measures between the bad pattern based on the bad pattern information and the accessed sensing data; and

calculating an error rate of the bad pattern based on the similarity measures,

wherein the calculating of the similarity measures comprises;

obtaining a first array of the accessed sensing data; and

calculating the similarity measures between a second array of the bad pattern information and the obtained first array,

wherein the obtaining further comprises;

dividing a time axis of the accessed sensing data into a predetermined number of sections, calculating representative values of measured values for each divided section, and storing the calculated representative values of the measured values;

normalizing the stored representative values using a mean and a variance of the stored representative values; and

converting the accessed sensing data into the first array by providing symbols allocated to each of the normalized representative values for each section.

6. The method of claim 5, wherein the calculating the similarity measures between a second array of the bad pattern information and the obtained first array comprises:

calculating a first symbol increasing/decreasing index array indicating whether a symbol value increases/decreases from an immediately previous section on a time-axis, for the first array of the accessed time series sensing data; and

calculating a similarity measure between a second symbol increasing/decreasing index array of the bad pattern information and the calculated first symbol increasing/decreasing index array.

7. The method of claim 6, wherein the calculating the similarity measures between the second array of the bad pattern information and the obtained first array further comprises:

calculating a similarity measure of an Euclidean ratio between the second array of the bad pattern information and the obtained first array; and

calculating a similarity measure of a correlation between the second array of the bad pattern information and the obtained first array.

8. An apparatus for verifying a bad pattern in sensing data, the apparatus comprising:

a verification parameter receiving unit configured to receive verification parameters including information on the bad pattern applied to sensing data of a sensor and information on a verification method;

a sensing data extracting unit configured to access the sensing data of each product, generated by the sensor based on verification method information;

a similarity measure calculating unit configured to calculate similarity measures between the bad pattern based on the bad pattern information and the sensing data, for each product;

a bad pattern verification unit configured to calculate an error rate of the bad pattern based on the similarity measures; and

a pre-processing unit configured to receive the sensing data of each product from the sensing data extracting unit, apply a pre-processing process to the sensing data of each product and supply the pre-processed data to the similarity measure calculator,

wherein the pre-processing unit comprises;

a sensor data division &

compression module configured to divide a time axis of the sensing data of each product into a predetermined number of sections, calculate representative values of sensing data for each divided section, and store the calculated representative values;

a normalization module configured to normalize the stored representative values using a mean and a variance of the stored representative values; and

a SAX conversion module SAX (Symbolic Aggregate approXimation) configured to convert the sensing data of each product into a first array by providing symbols allocated to each of the normalized representative values for each section and supplying the SAX converted data to the similarity measure calculator.

9. The apparatus of claim 8, wherein the pre-processing unit converts the sensing data based on the bad pattern information into the first array, and the similarity measure calculator calculates a first similarity measure, a second similarity measure and a third similarity measure between the first array of the bad pattern and a second array of the sensing data, for each product,

wherein each of the first to third similarity measures has a value in a range between 0, which means non-similarity, and 1, which means sameness, the first similarity measure being an Euclidean ratio between the bad pattern based on the bad pattern information and the second array of the sensing data, the second similarity measure being a correlation between the first array of the bad pattern and the second array of the sensing data, and the third similarity measure being a trend similarity between the first array of the bad pattern and the second array of the sensing data, and

wherein the trend similarity is a similarity measure between a first symbol increasing/decreasing index array, each item in the first symbol increasing/decreasing index array indicating whether each symbol value in the first symbol increasing/decreasing index array increases or decreases from the immediately previous section on a time-axis.

10. The apparatus of claim 9, wherein a determining information receiving unit queries information of each product, wherein the similarity measure calculator receives the information from the determining information receiving unit, calculates a number of products determined as good products among products having 1 as a value of at least one from among the first to the third similarity measures and calculates the error rate using the number of products determined as good products.

11. The apparatus of claim 8, wherein the sensing data is time series sensing data.

12. The apparatus of claim 8, wherein the accessing of the sensing data of each product, generated by the sensor occurs during a verification period.