# US 2015/0106655 A1 · Fault diagnosing method based on simulated vaccine

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2015/0106655 A1
- Document ID: US20150106655A1
- Status: Published application
- Kind: A1
- Application number: 14/235,062
- Filed: 2012-11-29
- Published: 2015-04-16
- Earliest claimed: 2012-07-03
- Assignees: Tsinghua University
- Applicants: Tsinghua University
- Inventors: Jingsong Zhao; Yiyang Dai; Yidan Shu; Jianfeng Zhu
- Web page: https://patentgrep.com/patent/US20150106655A1

## CPC

- G06F 11/1417: {Boot up procedures}
- G05B 17/02: electric
- G06F 30/20: Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30)
- G06F 2201/85: Active fault masking without idle spares
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)

## Abstract

A fault diagnosing method based on a simulated vaccine is provided. The fault diagnosing method comprises steps of: establishing a dynamic simulation model for simulating a start-up operation and a steady-state operation of a technological process according to process information, an operating procedure and historical data of the technological process; running the dynamic simulation model; extracting data within a first predetermined time from the simulated normal sample set to generate a normal simulated vaccine, and extracting data within a second predetermined time from the simulated fault sample set to generate a fault simulated vaccine; acquiring the historical data of the technological process; obtaining an immune antibody coefficient of the normal simulated vaccine, and generating a normal antibody library; and obtaining an immune antibody coefficient of the fault simulated vaccine, and generating a fault antibody library according to the fault simulated vaccine.

## Claims

### Claim 1

A fault diagnosing method based on a simulated vaccine, comprising steps of:
establishing a dynamic simulation model for simulating a start-up operation and a steady-state operation of a technological process according to process information, an operating procedure and historical data of the technological process;
running the dynamic simulation model to obtain a simulated sample set of simulated running data in a plurality of states, wherein the simulated sample set has various types of simulated samples and comprises a simulated normal sample set and a simulated fault sample set, the simulated normal sample set comprises simulated data sample sets obtained during a normal start-up operation and during a normal steady-state operation, and the simulated fault sample set comprises fault data sample sets corresponding to a plurality of faults generated during the start-up operation and during the steady-state operation;
extracting data within a first predetermined time from the simulated normal sample set to generate a normal simulated vaccine, and extracting data within a second predetermined time from the simulated fault sample set to generate a fault simulated vaccine;
acquiring the historical data of the technological process to generate a historical sample set comprising a historical normal sample set and a historical fault sample set;
obtaining an immune antibody coefficient of the normal simulated vaccine according to a historical normal sample in the historical normal sample set and a normal simulated vaccine corresponding to the historical normal sample in the historical normal sample set, and generating a normal antibody library according to the normal simulated vaccine; and
obtaining an immune antibody coefficient of the fault simulated vaccine according to a historical fault sample in the historical fault sample set and a fault simulated vaccine corresponding to the historical fault sample in the historical fault sample set, and generating a fault antibody library according to the fault simulated vaccine.

### Claim 2 (depends on 1)

The fault diagnosing method according to claim 1, further comprising:
calculating a minimum difference degree matrix between a fault start-up historical sample and a normal start-up historical sample by a DTW algorithm.

### Claim 3 (depends on 2)

The fault diagnosing method according to claim 2, further comprising:
running an actual technological process to obtain a sample of on-line running data; and
extracting a data fragment set within a predetermined time range before a current moment from the sample of the on-line running data.

### Claim 4 (depends on 3)

The fault diagnosing method according to claim 3, further comprising:
calculating a difference degree of an antigen at the current moment and each normal antibody in the normal antibody library to obtain a difference degree set; and
determining whether each difference degree in the difference degree set is less than a threshold of the normal antibody library, and if yes, determining that the technological process is normal, otherwise, determining that there is a fault in the technological process.

### Claim 5 (depends on 4)

The fault diagnosing method according to claim 4, further comprising:
determining a type of the fault in the technological process after determining that there is the fault in the technological process.

### Claim 6 (depends on 5)

The fault diagnosing method according to claim 5, further comprising:
for the start-up operation, acquiring normal start-up sample data of a normal start-up historical sample HNn×l under a condition similar to a current operating condition, and calculating a deviation matrix by the DTW algorithm or by a first predetermined formula as a diagnostic antigen;
for the steady-state operation, acquiring a value of each sampling point in a fragment of the on-line running data, and subtracting a datum D obtained at a moment when a fault is detected from the value of each sampling point to obtain a diagnostic antigen;
calculating a difference degree of a diagnostic antigen at the current moment and each fault antibody in the fault antibody library by the DTW algorithm or by a second predetermined formula; and
if the difference degree of the diagnostic antigen at the current moment and one fault antibody is less than a threshold of the fault antibody library, determining that the fault of the technological process is of a corresponding fault type.

### Claim 7 (depends on 1)

The fault diagnosing method according to claim 1, wherein the step of establishing the dynamic simulation model comprises:
establishing the dynamic simulation model by a process simulation software or by a differential equation.

### Claim 8 (depends on 1)

The fault diagnosing method according to claim 1, wherein the process information comprises a process flow chart, a material state, operating parameters, controller settings and environmental parameters.

### Claim 9 (depends on 1)

The fault diagnosing method according to claim 1, wherein the step of running the dynamic simulation model to obtain the simulated sample set of the simulated running data in the plurality of states further comprises:
normalizing each simulated datum in a set of the simulated running data by a following formula to obtain the simulated sample set:
X
=
0.5
+
x
-
X
_
X
max
-
X
min
where x is a simulated datum of a variable in a simulated sample, X is an average value of historical data of the variable x under an actual normal operating condition, and Xmax and Xmin are a maximum value and a minimum value of the historical data of the variable x under the actual normal operating condition respectively.

### Claim 10 (depends on 3)

The fault diagnosing method according to claim 3, wherein running the actual technological process to obtain the sample of the on-line running data further comprises:
normalizing the sample of the on-line running data.

### Claim 11 (depends on 1)

The fault diagnosing method according to claim 1, further comprising:
acquiring the historical data of the technological process; and
correcting the dynamic simulation model according to the historical data.

### Claim 12 (depends on 1)

The fault diagnosing method according to claim 1, wherein the step of running the dynamic simulation model to obtain the simulated sample set of the simulated running data in the plurality of states further comprises:
adding a disturbance within a normal operating range to the dynamic simulation model to generate the simulated normal sample set when the dynamic simulation model is run; and
adding a fault disturbance to the dynamic simulation model to generate the simulated fault sample set when the dynamic simulation model is run,
in which each of the disturbance within the normal operating range and the fault disturbance comprises environmental parameters, a noise and operating parameters.

### Claim 13 (depends on 6)

The fault diagnosing method according to claim 6, further comprising:
updating an antibody in a corresponding antibody library by the on-line running data according to a comparison between an on-line antigen and the normal antibody library and a comparison between the on-line antigen and the fault antibody library.
