Method for alerting of an event affecting a physical system

1 . A computer-implemented method for determining new occurrences of an event of a physical system to be manufactured in the course of designing a model of the physical on a Computer-Aided Design (CAD) system, the event being a cause of an incident of the physical system, the method comprising:

obtaining a first set and a third set of past events of the physical system, each past event comprising several attributes, at least one of the several attributes being a structured attribute and at least one of the several attributes being an unstructured attribute, each attribute being a symbol associated with data, the symbol characterizing a parameter of the physical system, the data associated with at least one attribute having been measured by a probe;

training a context sensitive auto-encoder with the third set of past events, the context sensitive auto-encoder being trained to take as input a vectorized event and to provide as output a signature of the event, the signature of the event being a compression of the event that contains as much as possible of the information necessary to reproduce the event;

obtaining a signature for each past event of the first set by:

vectorizing each past event, the vectorizing comprising converting unstructured attributes into a number of numerical attributes; and

applying the trained context sensitive auto-encoder to each vectorized past event;

obtaining a new event comprising several attributes, the new event being a cause of an incident of the physical system, data associated with at least one attribute of the new event having been measured by a probe;

computing a signature of the new event by:

vectorizing the new event; and

applying the trained context sensitive auto-encoder to the vectorized new event;

computing a similarity measure between the signature of the new event and each signature of each past event of the first set;

determining a plurality of past events closest to the new event, the plurality of past events having the highest similarity measures, the plurality of past events thereby forming a second set of past events, the number of the plurality of past events closest to the new event being predefined by a user;

computing a score of relevance for each structured attribute of each of the events of the second set, the computing comprising, for each structured attribute:

obtaining a distribution of values of the structured attribute on the second set;

obtaining a distribution of values of the structured attribute on the first set; and

comparing the distributions of values; and

generating a set of attributes by selecting the structured attributes of each of the events of the second set having the greater scores of relevance.

2 . The computer-implemented method of claim 1 , wherein the attributes of the second set for which the scores of relevance are computed are present in each event in the first set.

3 . The computer-implemented method of claim 2 , wherein the score of relevance of an attribute is computed by comparing distribution probabilities of values of the attribute on the second set with distribution probabilities of values of the attribute on the first set.

4 . The computer-implemented method of claim 1 , wherein a similarity metric used for determining the past events closest to the new event is one among:

cosine similarity;

Euclidian distance; and

inverse Euclidian distance.

5 . The computer-implemented method of claim 1 , further comprising, after forming a second set of past events:

computing at least one subset of past events of the second set; and

wherein computing the score of relevance further includes: computing the score of relevance for each attribute of in the at least one subset of past events of the second set.

6 . The computer-implemented method of claim 1 , wherein the past events of the first set and the new event are described by the same attributes.

7 . The computer-implemented method of claim 1 , wherein the new event and the signature of the new event are stored with the past events and the signatures of the past events thereby becoming a past event.

8 . A non-transitory computer readable medium having stored thereon a computer program comprising instructions for performing a method for determining new occurrences of an event of a physical system to be manufactured in the course of designing a model of the physical system on a Computer-Aided Design (CAD) system, the event being a cause of an incident of the physical system, the method comprising:

obtaining a first set and a third set of past events of the physical system, each past event comprising several attributes, at least one of the several attributes being a structured attribute and at least one of the several attributes being an unstructured attribute, each attribute being a symbol associated with data, the symbol characterizing a parameter of the physical system, the data associated with at least one attribute having been measured by a probe;

training a context sensitive auto-encoder with the third set of past events, the context sensitive auto-encoder being trained to take as input a vectorized event and to provide as output a signature of the event, the signature of the event being a compression of the event that contains as much as possible of the information necessary to reproduce the event;

obtaining a signature for each past event of the first set by:

vectorizing each past event, the vectorizing comprising converting unstructured attributes into a number of numerical attributes; and

applying the trained context sensitive auto-encoder to each vectorized past event;

obtaining a new event comprising several attributes, the new event being a cause of an incident of the physical system, data associated with at least one attribute of the new event having been measured by a probe;

computing a signature of the new event by:

vectorizing the new event; and

applying the trained context sensitive auto-encoder to the vectorized new event;

computing a similarity measure between the signature of the new event and each signature of each past event of the first set;

determining a plurality of past events closest to the new event, the plurality of past events having the highest similarity measures, the plurality of past events thereby forming a second set of past events, the number of the plurality of past events closest to the new event being predefined by a user;

computing a score of relevance for each structured attribute of each of the events of the second set, the computing comprising, for each structured attribute:

obtaining a distribution of values of the structured attribute on the second set;

obtaining a distribution of values of the structured attribute on the first set; and

comparing the distributions of values; and

generating a set of attributes by selecting the structured attributes of each of the events of the second set having the greater scores of relevance.

9 . A system comprising:

a processor coupled to a memory, the memory having recorded thereon a computer program for determining new occurrences of an event of a physical system to be manufactured in the course of designing a model of the physical system on a Computer-Aided Design (CAD) system, the event being a cause of an incident of the physical system, that when executed by the processor causes the processor to be configured to:

obtain a first set and a third set of past events of the physical system, each past event comprising several attributes, at least one of the several attributes being a structured attribute and at least one of the several attributes being an unstructured attribute, each attribute being a symbol associated with data, the symbol characterizing a parameter of the physical system, the data associated with at least one attribute having been measured by a probe;

train a context sensitive auto-encoder with the third set of past events, the context sensitive auto-encoder being trained to take as input a vectorized event and to provide as output a signature of the event, the signature of the event being a compression of the event that contains as much as possible of the information necessary to reproduce the event;

obtain a signature for each past event of the first set by the processor being configured to:

vectorize each past event, the vectorizing comprising converting unstructured attributes into a number of numerical attributes; and

apply the trained context sensitive auto-encoder to each vectorized past event;

obtain a new event comprising several attributes, the new event being a cause of an incident of the physical system, data associated with at least one attribute of the new event having been measured by a probe;

compute a signature of the new event by the processor being configured to:

vectorize the new event, and

apply the trained context sensitive auto-encoder to the vectorized new event;

compute a similarity measure between the signature of the new event and each signature of each past event of the first set;

determine a plurality of past events closest to the new event, the plurality of past events having the highest similarity measures, the plurality of past events thereby forming a second set of past events, the number of the plurality of past events closest to the new event being predefined by a user;

compute a score of relevance for each structured attribute of each of the events of the second set, the computing comprising, for each structured attribute, by the processor:

obtain a distribution of values of the structured attribute on the second set;

obtain a distribution of values of the structured attribute on the first set; and

comparing the distributions of values;

generate a set of attributes by selecting the structured attributes of each of the events of the second set having the greater scores of relevance.

10 . The computer-implemented method of claim 2 , wherein a similarity metric used for determining the past events closest to the new event is one among:

cosine similarity;

Euclidian distance; and

inverse Euclidian distance.

11 . The computer-implemented method of claim 3 , wherein a similarity metric used for determining the past events closest to the new event is one among:

cosine similarity;

Euclidian distance; and

inverse Euclidian distance.

12 . The computer implemented method of claim 1 , wherein the comparing of the distributions of values comprises a two-sample Kolmogorov Smirnov test.

13 . The computer implemented method of claim 1 , further comprising:

suggesting to a user a value of at least one of the selected structured attributes of each of the past events that are characteristic of a number of incidents, the suggested value minimizing the chances of an occurrence of the incident associated to the new event; and

selecting, by the user, the suggested value, thereby reducing the defects in the lifetime of the physical system to be manufactured.