The sampling of the training data is a bottleneck in the development of artificial intelligence(AI)models due to the processing of huge amounts of data or to the difficulty of access to the data in industrial practice...The sampling of the training data is a bottleneck in the development of artificial intelligence(AI)models due to the processing of huge amounts of data or to the difficulty of access to the data in industrial practices.Active learning(AL)approaches are useful in such a context since they maximize the performance of the trained model while minimizing the number of training samples.Such smart sampling methodologies iteratively sample the points that should be labeled and added to the training set based on their informativeness and pertinence.To judge the relevance of a data instance,query rules are defined.In this paper,we propose an AL methodology based on a physics-based query rule.Given some industrial objectives from the physical process where the AI model is implied in,the physics-based AL approach iteratively converges to the data instances fulfilling those objectives while sampling training points.Therefore,the trained surrogate model is accurate where the potentially interesting data instances from the industrial point of view are,while coarse everywhere else where the data instances are of no interest in the industrial context studied.展开更多
Describing the orientation state of the particles is often critical in fibre suspension applications.Macroscopic descriptors,the so-called second-order orientation tensor(or moment)leading the way,are often preferred ...Describing the orientation state of the particles is often critical in fibre suspension applications.Macroscopic descriptors,the so-called second-order orientation tensor(or moment)leading the way,are often preferred due to their low computational cost.Closure problems however arise when evolution equations for the moments are derived from the orientation distribution functions and the impact of the chosen closure is often unpredictable.In this work,our aim is to provide macroscopic simulations of orientation that are cheap,accurate and closure-free.To this end,we propose an innovative data-based approach to the upscaling of orientation kinematics in the context of fibre suspensions.Since the physics at the microscopic scale can be modelled reasonably enough,the idea is to conduct accurate offline direct numerical simulations at that scale and to extract the corresponding macroscopic descriptors in order to build a database of scenarios.During the online stage,the macroscopic descriptors can then be updated quickly by combining adequately the items from the database instead of relying on an imprecise macroscopic model.This methodology is presented in the well-known case of dilute fibre suspensions(where it can be compared against closure-based macroscopic models)and in the case of suspensions of confined or electrically-charged fibres,for which state-of-the-art closures proved to be inadequate or simply do not exist.展开更多
文摘The sampling of the training data is a bottleneck in the development of artificial intelligence(AI)models due to the processing of huge amounts of data or to the difficulty of access to the data in industrial practices.Active learning(AL)approaches are useful in such a context since they maximize the performance of the trained model while minimizing the number of training samples.Such smart sampling methodologies iteratively sample the points that should be labeled and added to the training set based on their informativeness and pertinence.To judge the relevance of a data instance,query rules are defined.In this paper,we propose an AL methodology based on a physics-based query rule.Given some industrial objectives from the physical process where the AI model is implied in,the physics-based AL approach iteratively converges to the data instances fulfilling those objectives while sampling training points.Therefore,the trained surrogate model is accurate where the potentially interesting data instances from the industrial point of view are,while coarse everywhere else where the data instances are of no interest in the industrial context studied.
文摘Describing the orientation state of the particles is often critical in fibre suspension applications.Macroscopic descriptors,the so-called second-order orientation tensor(or moment)leading the way,are often preferred due to their low computational cost.Closure problems however arise when evolution equations for the moments are derived from the orientation distribution functions and the impact of the chosen closure is often unpredictable.In this work,our aim is to provide macroscopic simulations of orientation that are cheap,accurate and closure-free.To this end,we propose an innovative data-based approach to the upscaling of orientation kinematics in the context of fibre suspensions.Since the physics at the microscopic scale can be modelled reasonably enough,the idea is to conduct accurate offline direct numerical simulations at that scale and to extract the corresponding macroscopic descriptors in order to build a database of scenarios.During the online stage,the macroscopic descriptors can then be updated quickly by combining adequately the items from the database instead of relying on an imprecise macroscopic model.This methodology is presented in the well-known case of dilute fibre suspensions(where it can be compared against closure-based macroscopic models)and in the case of suspensions of confined or electrically-charged fibres,for which state-of-the-art closures proved to be inadequate or simply do not exist.
