3. Producción
Browse
3 results
Search Results
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Fault diagnosis via neural ordinary differential equations(MDPI AG, 2021-05-01)Implementation of model-based fault diagnosis systems can be a difficult task due to the complex dynamics of most systems, an appealing alternative to avoiding modeling is to use machine learning-based techniques for which the implementation is more affordable nowadays. However, the latter approach often requires extensive data processing. In this paper, a hybrid approach using recent developments in neural ordinary differential equations is proposed. This approach enables us to combine a natural deep learning technique with an estimated model of the system, making the training simpler and more efficient. For evaluation of this methodology, a nonlinear benchmark system is used by simulation of faults in actuators, sensors, and process. Simulation results show that the proposed methodology requires less processing for the training in comparison with conventional machine learning approaches since the data-set is directly taken from the measurements and inputs. Furthermore, since the model used in the essay is only a structural approximation of the plant; no advanced modeling is required. This approach can also alleviate some pitfalls of training data-series, such as complicated data augmentation methodologies and the necessity for big amounts of data. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Fault detection and isolation for UAVs using neural ordinary differential equations(Elsevier B.V., 2022-01-01)In recent years, the increasing complexity and diversity of data-based fault detection and isolation (FDI) methods usually require high computational efforts in the pre-processing stage, large amounts of data, and, most of the time, some feature extraction to obtain relevant information for the data-based algorithms. This paper proposes using the Neural Ordinary Differential Equations (NODE) framework to represent the dynamics of the studied plant and later employ such representation in FDI system design. Such an approach enables loss optimization to be performed jointly in the plant dynamics and external inputs without previous use of complex pre-processing and is useful for working with nonlinear systems. The approach is first validated using a simulated Unmanned Aerial Vehicle (UAV) and later applied to a data-set that contains actuators and sensors faults. Ultimately, the proposed approach is compared with other usual machine learning techniques, showing better performance metrics. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Flotation Process Fault Detection and Isolation Using Neural ODE for Generation of Vector-Field Features(Elsevier B.V., 2023-07-01)Flotation in the mining industry is of vital importance for obtaining the right quality of product with efficiency and represents a critical process where possible failures must be monitored at all times. In this paper, complete fault detection and isolation system (FDI) based on the Neural Ordinary Differential Equations (NODE) framework is proposed; the NODE is employed to represent the dynamics of the studied plant based on the measured variables and inputs. Then, a classifier can be used to identify the faults based on the projections of the derivatives or local vector field generated by the NODE using the estimations and actual measurements. The proposed approach is applied to a controlled mining flotation process that has perturbations. The solution is compared with other known machine learning techniques showing better performance metrics. Moreover, it is demonstrated with t-SNE representation that features generated from the NODE model improve the classification.