Diagnosing method of engine condition and diagnostic modeling method thereof

1. A method of diagnosing an engine condition, comprising:

measuring a vibration of an engine, wherein assembly of the engine is completed in an automated end of line (EOL) process of an engine production; and

primarily diagnosing the vibration of the engine during the EOL process using a primary deep learning classification model in which vibration signals of the engine are classified according to a feature through learning using a plurality of algorithms on the basis of big data with respect to multiple vibration signals measured at three or more positions on the engine,

wherein the assembled state of the engine in the EOL process is classified into a normal state or an abnormal state by performing the primary diagnosing of the vibration of the engine; and

wherein the primary deep learning classification model is constructed by sequentially learning the multiple vibration signals using a recursive neural network (RNN) (many-to-one) based classification algorithm and a convolution neural network (CNN) and attention mechanism algorithm;

when the engine in the EOL process is classified as being abnormal in the primary diagnosis, secondarily diagnosing the vibration of the engine in the EOL process using a secondary deep learning diagnostic model in which a diagnostic reference setting value is determined according to an abnormal case on the basis of the big data with respect to the multiple vibration signals.

2. The method of claim 1 , wherein the measuring the vibration of the engine in the EOL process includes measuring a vibration at positions that are the same as the measurement positions of the multiple vibration signals for the primary deep learning classification model.

3. The method of claim 1 , wherein the big data with respect to the multiple vibration signals for constructing the primary deep learning classification model includes data with respect to a normal signal and plural types of abnormal vibration signals according to an acceleration or constant speed driving condition.

4. The method of claim 3 , wherein each of the multiple vibration signals includes data over an entire frequency range.

5. The method of claim 1 , wherein the measurement of the vibration of the engine in the EOL stage is re-performed on the engine which is repaired according to the abnormal case diagnosed in the secondary diagnosis.

6. The method of claim 1 , further comprising:

storing diagnosis results, which are obtained in the primary diagnosis and the secondary diagnosis, in a central server; and

including the data, which is stored in the central server, and vibration data for the engine, which is separately measured, in the big data.

7. The method of claim 1 , wherein the second deep learning diagnostic model is constructed by learning the multiple vibration signals using frame energy, a pitch, zero crossing, binary classification, and an ensemble deep neural network (DNN)/result selection algorithm.

8. A diagnostic modeling method of an engine condition, comprising:

collecting big data with respect to a vibration signal of an engine;

measuring multiple vibration signals at three or more positions on the engine in an automated end of line (EOL) process of an engine production;

analyzing data and extracting a feature for supervised learning on the basis of the big data and the multiple vibration signals;

constructing a primary deep learning classification model in which the vibration signal of the engine is classified according to a feature through learning using a plurality of algorithms on the basis of the big data; and

constructing a secondary deep learning diagnostic model in which a diagnostic reference setting value is determined according to an abnormal case classified using the primary deep learning classification model on the basis of the big data;

wherein the primary deep learning classification model is constructed by sequentially learning the multiple vibration signals using a recursive neural network (RNN) (many-to-one) based classification algorithm, a convolution neural network (CNN) and attention mechanism algorithm.

9. The method of claim 8 , wherein each of classification accuracy of the primary deep learning classification model and diagnostic accuracy of the secondary deep learning diagnostic model is 99.9% or more.

10. The method of claim 9 , wherein, when the classification accuracy of the primary deep learning classification model or the diagnostic accuracy of the secondary deep learning diagnostic model is less than 99.9%, the analyzing of the data and the extracting of the feature for the supervised learning are re-performed.

11. The method of claim 8 , wherein the big data with respect to the multiple vibration signals for constructing the primary deep learning classification model includes data with respect to a normal signal and plural types of abnormal vibration signals according to an acceleration or constant speed driving condition.

12. The method of claim 11 , wherein each of the multiple vibration signals includes data over an entire frequency range.

13. The method of claim 8 , wherein the second deep learning diagnostic model is constructed by learning the multiple vibration signals using frame energy, a pitch, zero crossing, binary classification, and an ensemble deep neural network (DNN)/result selection algorithm.