-
An Outline of Prognostics and Health Management Large Model: Concepts, Paradigms, and Challenges
Authors:
Laifa Tao,
Shangyu Li,
Haifei Liu,
Qixuan Huang,
Liang Ma,
Guoao Ning,
Yiling Chen,
Yunlong Wu,
Bin Li,
Weiwei Zhang,
Zhengduo Zhao,
Wenchao Zhan,
Wenyan Cao,
Chao Wang,
Hongmei Liu,
Jian Ma,
Mingliang Suo,
Yujie Cheng,
Yu Ding,
Dengwei Song,
Chen Lu
Abstract:
Prognosis and Health Management (PHM), critical for ensuring task completion by complex systems and preventing unexpected failures, is widely adopted in aerospace, manufacturing, maritime, rail, energy, etc. However, PHM's development is constrained by bottlenecks like generalization, interpretation and verification abilities. Presently, generative artificial intelligence (AI), represented by Larg…
▽ More
Prognosis and Health Management (PHM), critical for ensuring task completion by complex systems and preventing unexpected failures, is widely adopted in aerospace, manufacturing, maritime, rail, energy, etc. However, PHM's development is constrained by bottlenecks like generalization, interpretation and verification abilities. Presently, generative artificial intelligence (AI), represented by Large Model, heralds a technological revolution with the potential to fundamentally reshape traditional technological fields and human production methods. Its capabilities, including strong generalization, reasoning, and generative attributes, present opportunities to address PHM's bottlenecks. To this end, based on a systematic analysis of the current challenges and bottlenecks in PHM, as well as the research status and advantages of Large Model, we propose a novel concept and three progressive paradigms of Prognosis and Health Management Large Model (PHM-LM) through the integration of the Large Model with PHM. Subsequently, we provide feasible technical approaches for PHM-LM to bolster PHM's core capabilities within the framework of the three paradigms. Moreover, to address core issues confronting PHM, we discuss a series of technical challenges of PHM-LM throughout the entire process of construction and application. This comprehensive effort offers a holistic PHM-LM technical framework, and provides avenues for new PHM technologies, methodologies, tools, platforms and applications, which also potentially innovates design, research & development, verification and application mode of PHM. And furthermore, a new generation of PHM with AI will also capably be realized, i.e., from custom to generalized, from discriminative to generative, and from theoretical conditions to practical applications.
△ Less
Submitted 1 July, 2024;
originally announced July 2024.
-
Variation of Critical Crystallization Pressure for the Formation of Square Ice in Graphene Nanocapillaries
Authors:
Zhen Zeng,
Kai Sun,
Rui Chen,
Mengshan Suo,
Zhizhao Che,
Tianyou Wang
Abstract:
Two-dimensional square ice in graphene nanocapillaries at room temperature is a fascinating phenomenon and has been confirmed experimentally. Instead of temperature for bulk ice, the high van der Waals pressure becomes an all-important factor to induce the formation of square ice and needs to be studied further. By all-atom molecular dynamics simulations of water confined between two parallel grap…
▽ More
Two-dimensional square ice in graphene nanocapillaries at room temperature is a fascinating phenomenon and has been confirmed experimentally. Instead of temperature for bulk ice, the high van der Waals pressure becomes an all-important factor to induce the formation of square ice and needs to be studied further. By all-atom molecular dynamics simulations of water confined between two parallel graphene sheets, which are changed in size (the length and the width of the graphene sheets) over a wide range, we find that the critical crystallization pressure for the formation of square ice in nanocapillary strongly depends on the size of the graphene sheet. The critical crystallization pressure slowly decreases as the graphene size increases, converging to approximately macroscopic crystallization pressure. The unfreezable threshold for graphene size is obtained by estimating the actual pressure and it is difficult to form the square ice spontaneously in practice when the graphene sheet is smaller than the threshold. Moreover, the critical crystallization pressure fluctuates when the graphene size is minuscule, and the range of oscillation narrows as the sheet size increases, converging to the macroscopic behavior of a single critical icing pressure for large sheets. The graphene size also affects the stability and crystallization time of square ice.
△ Less
Submitted 7 July, 2023;
originally announced July 2023.
-
Two-dimensional partitioned square ice confined in graphene/graphite nanocapillaries
Authors:
Zhen Zeng,
Tianyou Wang,
Rui Chen,
Mengshan Suo,
Kai Sun,
Panagiotis E. Theodorakis,
Zhizhao Che
Abstract:
As one of the most fascinating confined water/ice phenomena, two-dimensional square ice has been extensively studied and experimentally confirmed in recent years. Apart from the unidirectional homogeneous square icing patterns considered in previous studies, the multidirectional partitioned square icing patterns are discovered in this study and characterized by molecular dynamics (MD) simulations.…
▽ More
As one of the most fascinating confined water/ice phenomena, two-dimensional square ice has been extensively studied and experimentally confirmed in recent years. Apart from the unidirectional homogeneous square icing patterns considered in previous studies, the multidirectional partitioned square icing patterns are discovered in this study and characterized by molecular dynamics (MD) simulations. Square icing parameters are proposed to quantitatively distinguish the partitioned patterns from the homogeneous patterns and the liquid water. The number of graphene monolayers n is varied in this study, and the results show that it is more energetically favorable to form partitioned square icing patterns when the water molecules are confined between graphite sheets (n >= 2) compared to graphene (n = 1). This phenomenon is insensitive to n as long as n >= 2, because of the short-range nature of the interaction between water molecules and the carbon substrate. Moreover, it is energetically unfavorable to form partitioned square icing patterns for a single layer of water molecules even for n >= 2, verifying that the interaction between layers of water molecules is another dominant factor in the formation of partitioned structures. The conversion from partitioned structure to homogenous square patterns is investigated by changing the pressure and the temperature. Based on the comprehensive MD simulations, this study unveils the formation mechanism of the partitioned square icing patterns.
△ Less
Submitted 7 April, 2022;
originally announced April 2022.
-
Using Python for Model Inference in Deep Learning
Authors:
Zachary DeVito,
Jason Ansel,
Will Constable,
Michael Suo,
Ailing Zhang,
Kim Hazelwood
Abstract:
Python has become the de-facto language for training deep neural networks, coupling a large suite of scientific computing libraries with efficient libraries for tensor computation such as PyTorch or TensorFlow. However, when models are used for inference they are typically extracted from Python as TensorFlow graphs or TorchScript programs in order to meet performance and packaging constraints. The…
▽ More
Python has become the de-facto language for training deep neural networks, coupling a large suite of scientific computing libraries with efficient libraries for tensor computation such as PyTorch or TensorFlow. However, when models are used for inference they are typically extracted from Python as TensorFlow graphs or TorchScript programs in order to meet performance and packaging constraints. The extraction process can be time consuming, impeding fast prototy**. We show how it is possible to meet these performance and packaging constraints while performing inference in Python. In particular, we present a way of using multiple Python interpreters within a single process to achieve scalable inference and describe a new container format for models that contains both native Python code and data. This approach simplifies the model deployment story by eliminating the model extraction step, and makes it easier to integrate existing performance-enhancing Python libraries. We evaluate our design on a suite of popular PyTorch models on Github, showing how they can be packaged in our inference format, and comparing their performance to TorchScript. For larger models, our packaged Python models perform the same as TorchScript, and for smaller models where there is some Python overhead, our multi-interpreter approach ensures inference is still scalable.
△ Less
Submitted 1 April, 2021;
originally announced April 2021.
