Supervision method and system for predicting defects during additive manufacturing processes

1 . A method for supervising an additive manufacturing process performed by an additive manufacturing device, said method comprising:

receiving a set of data by a supervision system, said set of data comprising product geometry data, production data, production history data, and measurements from said additive manufacturing device, wherein the production history data comprises organized data from past manufactured products, associating for said past manufactured products, product geometry data, production data, measurements and information relating to defects determined by the supervision system during the additive manufacturing process; and,

iteratively performing at said supervision system, an iteration corresponding to measurements related to a past timeframe fed to the supervision system;

predicting an occurrence of a defect comprising predicting a future occurrence of a defect within a future timeframe from said set of data, based on a predictive model tuned with a supervised training dataset comprising manual labels specifying whether a defect has happened or will happen within a given timeframe;

when the occurrence of the defect is predicted, determining a type of defect, based on a predictive model tuned with a supervised training dataset comprising manual labels specifying the type of defect that has happened or will happen within a given timeframe; and

deciding for stopping or not said additive manufacturing process according to said type of defect and of a seriousness level for said type of defect, using a predictive model based on production history constraints.

2 . The method according to claim 1 , further comprising triggering at least one correction action according to said type of defect, said at least one correction action comprising a command transmitted to said additive manufacturing device.

3 . The method according to claim 2 , wherein said command replaces an initial command from an additive manufacturing pilot module intercepted by said supervision system.

4 . The method according to claim 2 , wherein triggering at least one correction action is realized by a predictive model selected according to said type of defect.

5 . The method according to claim 1 , further comprising a preliminary step for detecting defects before starting said additive manufacturing process according to a subset of said set of data comprising product geometry data, production data, production history data.

6 . The method according to claim 5 , wherein predicting the occurrence of a defect comprises predicting a future occurrence of a defect.

7 . A non-transitory computer readable medium encoding a machine-executable program of instructions to perform a method according to claim 1 .

8 . A supervision system for supervising an additive manufacturing process performed by an additive manufacturing device, comprising:

an interface for receiving a set of data, said set of data comprising product geometry data, production data, production history data, and measurements from said additive manufacturing device wherein the production history data comprises organized data from past manufactured products, associating for said past manufactured products, product geometry data, production data, measurements and information relating to defects determined by the supervision system during the additive manufacturing process;

a processor and a memory storing instructions that, when executed by the processor, cause the supervision system to iteratively perform, an iteration corresponding to measurements related to a past timeframe fed to the supervision system:

for a prediction module configured to predict an occurrence of a defect configured to predict a future occurrence of a defect within a future timeframe from said set of data based on a predictive model tuned with a supervised training dataset comprising manual labels specifying whether a defect has happened or will happen within a given timeframe; and

a defect determination module configured to, when the occurrence of the defect is predicted, determine a type of defect, based on a predictive model tuned with a supervised training dataset comprising manual labels specifying the type of defect that has happened or will happen within a given timeframe; and

a decision module configured to decide for stopping or not said additive manufacturing process according to said type of defect and of a seriousness level for said type of defect, using a predictive model based on production history constraints.

9 . A system comprising:

an additive manufacturing device; and

a supervision system according to claim 8 configured to supervise an additive manufacturing process performed by said additive manufacturing device.