AS
A.A. Schouten
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
2 records found
1
Leveraging Autoencoders
To Enhance Model Order Reduction for Non-linear Mechanical Dynamical Systems
The computational demands of finite element simulations, particularly in predicting the time-dependent response of high-dimensional non-linear dynamical systems, pose significant challenges. To overcome these challenges, researchers have developed model order reduction (MOR) methods, which aim to reduce computational complexity by utilizing lower-dimensional models. This thesis proposes a MOR technique that simultaneously learns both the projection to, and the reduced dynamics on, a lower-dimensional manifold using autoencoders, a type of neural network. During training, the known linear part of the reduced dynamics is used to aid the optimization process, leading to an effective method of simultaneous projection and linear informed training (SPLIT). SPLIT demonstrates outstanding performance on the test case of a 2D cantilever beam, and is capable of making non-linear forced response predictions, even though being trained on unforced decaying trajectories. Even in scenarios involving highly non-linear behaviour, such as when the beam folds over itself, SPLIT continues to make accurate predictions, while other MOR techniques fail. This work highlights the potential of autoencoders to advance the field of MOR and improve the efficiency and reliability of simulations for complex dynamical systems.
...
The computational demands of finite element simulations, particularly in predicting the time-dependent response of high-dimensional non-linear dynamical systems, pose significant challenges. To overcome these challenges, researchers have developed model order reduction (MOR) methods, which aim to reduce computational complexity by utilizing lower-dimensional models. This thesis proposes a MOR technique that simultaneously learns both the projection to, and the reduced dynamics on, a lower-dimensional manifold using autoencoders, a type of neural network. During training, the known linear part of the reduced dynamics is used to aid the optimization process, leading to an effective method of simultaneous projection and linear informed training (SPLIT). SPLIT demonstrates outstanding performance on the test case of a 2D cantilever beam, and is capable of making non-linear forced response predictions, even though being trained on unforced decaying trajectories. Even in scenarios involving highly non-linear behaviour, such as when the beam folds over itself, SPLIT continues to make accurate predictions, while other MOR techniques fail. This work highlights the potential of autoencoders to advance the field of MOR and improve the efficiency and reliability of simulations for complex dynamical systems.
Context: In astronomy, collision detection is the computational problem of finding when planets, asteroids or satellites collide. Computer simulations make it possible to show very precisely how bodies move through space, which makes it possible to detect collisions. However, these simulations require a lot of computing power if there are many bodies, which is of course undesirable.
Aims: The aim of this report is to provide more insight into a collision detection method that does not use the computationally expensive simulations, but rather fast mathematical techniques.
Methods: It can be shown that the collision detection problem between two planets is equivalent to finding the integer point (𝑘, 𝑙) between two parallel lines, closest to the origin. The solution uses the continued fraction of the ratio of the orbital periods of the planets, as the slopes of the lines are this ratio. The convergents of the continued fraction form basis vectors with which the point (𝑘, 𝑙) is searched for. This is called lattice basis reduction.
Results: Where brute force always takes 𝑘 steps to find the integer point (𝑘, 𝑙), the method with continued fractions and lattice basis reduction often finds the point in about √𝑘 steps. The worst case however is still 𝑘 steps, the same as brute force. This only happens if there are no basis transformations, so luckily 𝑘 is then often small.
Conclusions: When there are many bodies between which collisions need to be detected, lattice basis reduction provides a fast method. For the basis vectors, the convergents of the continued fraction of the ratio of the orbital periods of the planets must be used. If the solution cannot be reached exactly, (𝑘, 𝑙) can also be estimated.
...
Aims: The aim of this report is to provide more insight into a collision detection method that does not use the computationally expensive simulations, but rather fast mathematical techniques.
Methods: It can be shown that the collision detection problem between two planets is equivalent to finding the integer point (𝑘, 𝑙) between two parallel lines, closest to the origin. The solution uses the continued fraction of the ratio of the orbital periods of the planets, as the slopes of the lines are this ratio. The convergents of the continued fraction form basis vectors with which the point (𝑘, 𝑙) is searched for. This is called lattice basis reduction.
Results: Where brute force always takes 𝑘 steps to find the integer point (𝑘, 𝑙), the method with continued fractions and lattice basis reduction often finds the point in about √𝑘 steps. The worst case however is still 𝑘 steps, the same as brute force. This only happens if there are no basis transformations, so luckily 𝑘 is then often small.
Conclusions: When there are many bodies between which collisions need to be detected, lattice basis reduction provides a fast method. For the basis vectors, the convergents of the continued fraction of the ratio of the orbital periods of the planets must be used. If the solution cannot be reached exactly, (𝑘, 𝑙) can also be estimated.
...
Context: In astronomy, collision detection is the computational problem of finding when planets, asteroids or satellites collide. Computer simulations make it possible to show very precisely how bodies move through space, which makes it possible to detect collisions. However, these simulations require a lot of computing power if there are many bodies, which is of course undesirable.
Aims: The aim of this report is to provide more insight into a collision detection method that does not use the computationally expensive simulations, but rather fast mathematical techniques.
Methods: It can be shown that the collision detection problem between two planets is equivalent to finding the integer point (𝑘, 𝑙) between two parallel lines, closest to the origin. The solution uses the continued fraction of the ratio of the orbital periods of the planets, as the slopes of the lines are this ratio. The convergents of the continued fraction form basis vectors with which the point (𝑘, 𝑙) is searched for. This is called lattice basis reduction.
Results: Where brute force always takes 𝑘 steps to find the integer point (𝑘, 𝑙), the method with continued fractions and lattice basis reduction often finds the point in about √𝑘 steps. The worst case however is still 𝑘 steps, the same as brute force. This only happens if there are no basis transformations, so luckily 𝑘 is then often small.
Conclusions: When there are many bodies between which collisions need to be detected, lattice basis reduction provides a fast method. For the basis vectors, the convergents of the continued fraction of the ratio of the orbital periods of the planets must be used. If the solution cannot be reached exactly, (𝑘, 𝑙) can also be estimated.
Aims: The aim of this report is to provide more insight into a collision detection method that does not use the computationally expensive simulations, but rather fast mathematical techniques.
Methods: It can be shown that the collision detection problem between two planets is equivalent to finding the integer point (𝑘, 𝑙) between two parallel lines, closest to the origin. The solution uses the continued fraction of the ratio of the orbital periods of the planets, as the slopes of the lines are this ratio. The convergents of the continued fraction form basis vectors with which the point (𝑘, 𝑙) is searched for. This is called lattice basis reduction.
Results: Where brute force always takes 𝑘 steps to find the integer point (𝑘, 𝑙), the method with continued fractions and lattice basis reduction often finds the point in about √𝑘 steps. The worst case however is still 𝑘 steps, the same as brute force. This only happens if there are no basis transformations, so luckily 𝑘 is then often small.
Conclusions: When there are many bodies between which collisions need to be detected, lattice basis reduction provides a fast method. For the basis vectors, the convergents of the continued fraction of the ratio of the orbital periods of the planets must be used. If the solution cannot be reached exactly, (𝑘, 𝑙) can also be estimated.