ISEL - Eng. Mecan. - Comunicações
Permanent URI for this collection
Browse
Browsing ISEL - Eng. Mecan. - Comunicações by Issue Date
Now showing 1 - 10 of 162
Results Per Page
Sort Options
- Industrial actuator diagnosis using hierarchical fuzzy neural networksPublication . Mendes, Mário J. G. C.; Calado, João Manuel Ferreira; Sousa, J. M. C.; Costa, J. M. G. Sá daIn this paper a hierarchical structure offuzzy neural networks (FNNs) and how to train it for fault isolation given an appropriate data patterns, are presented. Fault symptoms concerning multiple simultaneous faults are harder to learn than those associated with single faults. Furthermore, the larger the set of faults, the larger the set of fault symptoms will be and, hence, the longer and less certain the training outcome. In order to overcome this problem, the proposed approach has a hierarchical structure of three levels where several FNNs are used. Thus, a large number ofpatterns are divided into many smaller subsets so that the classification can be carried out more efficiently. One ofthe advantages of this approach is that multiple faults can be detected in new data even ifthe network is trained only with datarepresenting single abrupt faults. A continuous binary distillation column having several actuated valves with PID loops has been used as testbed for the proposed approach.
- Neuro-fuzzy techniques in FDI system for sugar factory actuatorsPublication . Mendes, Mário J. G. C.; Kowal, Marek; Korbicz, Józef; Costa, José M. G. Sá daFault diagnosis systems have an important role in industrial plants because the early fault detection and isolation (FDI) can minimize damages in the plants. The main aim of this work is to propose a two-stage neuro-fuzzy approach as a fault diagnosis system in dynamic processes. The first stage of the system is responsible for fault detection and is implemented using a neuro-fuzzy model. The second stage of the system is responsible for fault isolation and is built using an hierarchical structure of fuzzy neural networks. The FDI system is applied to fault diagnosis in the sugar factory actuators.
- Fault isolation approach using a PROFIBUS network: a case studyPublication . Mendes, Mário J. G. C.; Kowal, Marek; Calado, João Manuel Ferreira; Korbicz, Józef; Costa, J. M. G. Sá daThis paper presents the second stage, of a two-stage neuro-fuzzy system, used for fault isolation (FI) in dynamic processes and it`s built using a hierarchical structure of fuzzy neural networks. The current approach is tested under a hardware bench constructed with componentes commonly used in the industry and consists on a pilot plant under supervision, a supervision unit, a fault detection and isolation unit and a fault simulation unit. All the elements are connected to a PROFIBUS network, which acts as the communication system for exchanging information between the automation system and the distributed field devices.
- Pruning algorithm applied to a hierarchical structure of fuzzy neural networks: case studyPublication . Mendes, Mário J. G. C.; Calado, João Manuel Ferreira; Costa, J. M. G. Sá daThis research paper is concerned with the fault detection and isolation (FDI) problem, or more exactly, with a hierarchical structure of fuzzy neural networks (HFNN) used for fault isolation purposes in industrial processes. The main aim of this research work is to optimise the number of neurons in the hidden layer of all fuzzy neural networks (FNNs) used in the HFNN. Thus, the optimal brain surgeon (OBS) pruning algorithm has been used to prune all FNNs. After the OBS optimisation, the HFNN structure continues to be able to isolate correctly, abrupt and incipient, single and multiple faults. At the same time, the structure became simpler and better generalisation capabilities have been observed. A continuous binary distillation column having several actuated valves with PID control loops has been used as test bed of the proposed approach.
- “State-of-art” da detecção e diagnóstico de avarias em processos industriaisPublication . Mendes, Mário J. G. C.Ao longo das últimas décadas tem havido um interesse crescente, por parte da indústria e da comunidade académica, pelos sistemas de detecção e diagnóstico de avarias (sistemas FDI). A crescente complexidade dos processos e do seu controlo, assim como a necessidade de tornar os processos mais seguros e fiáveis incentivou a procura e desenvolvimento de sistemas FDI. O desenvolvimento e melhoria deste tipo de sistemas seguiu de perto a evolução na tecnologia, nomeadamente a evolução informática, que permitiu maior facilidade na análise de grandes quantidades de dados e aplicação de sistemas FDI em tempo real. Este artigo pretende fazer um ponto da situação, a nível mundial, desta área de investigação.
- Neuro-fuzzy fault detection approach using a Profibus networkPublication . Calado, João Manuel Ferreira; Mendes, Mário J. G. C.; Kowal, Marek; Korbicz, Józef; Costa, J. M. G. Sá daGenerally three methodologies to develop and test FDI algorithms can de distinguished: software benches, hardware benches and industrial data. The current approach uses a hardware bench constructed with components commonly used in industry that consists on a pilot plant under supervision, a supervision unit, a fault detection unit and a fault simulation unit. All elements are connected to a PROFIBUS network that acts as the communication system exchanging information between automation system and distributed field devices. A fault detection methodology, which is based on neuro-fuzzy models, has been developed and implemented. During the current studies actuator faults, sensor faults and leakages have been considered as incipient and abrupt faults. Several studies have also been performed under multiple simultaneous faulty scenarios.
- Fault detection scheme using the agents paradigmPublication . Mendes, Mário J. G. C.; Calado, João Manuel Ferreira; Costa, J. M. G. Sá daAn agent based fault detection (FD) system for complexa nd dynamic processes is proposed in this work. The system is based in the agent paradigm where the modularity and complexity of the processes are important aspects in the FD system constructed. In the future, the FD agents must be able to cooperate and communicate with other systems to achieve a satisfactory performance, as a part of a fault tolerant control multi-agent system. The FD agents proposed here have hybrid architectures based in a horizontal layered architecture. Two types of FD agents are proposed, one based in decomposition wavelet methods with limit checking and other based in neural networks ARX models for residual generation. The agent based FD scheme proposed is applied in a three tank process.
- FDI/FTC for complex networked control systems based on multi-agentsPublication . Costa, José Sá da; Mendes, Mário J. G. C.When dealing with large-scale complex networked control systems, designing FDI/FTC systems is a very difficult task due to the large number of sensors and actuators spatially distributed and networked connected. Any solution given to this problem must take into account that practitioners prefer rather simplistic solutions since in practice, simple and verifiable principles always win the competition versus complex solutions that are usually characterized by instability, unpredictable behaviour and large computational burden. The FDI/FTC framework presented in this paper is able to achieve this goal by using simple and verifiable principles coming mainly from a decentralized design based on causal modelling partitioning of the NCS and distributed computing using multiagents systems, allowing the use of well established FDI/FTC methodologies or new ones developed taking into account the NCS specificities.
- Control of the archimedes wave swing using neural networksPublication . Beirão, Pedro; Mendes, Mário J. G. C.; Valério, Duarte; Costa, José Sá daThis paper addresses the control of the Archimedes Wave Swing, a fully-submerged Wave Energy Converter (WEC), of which a prototype has already been built and tested. Simulation results are presented in which Internal Model Control (IMC) is used, both with linear models and with non-linear neural network (NN) models. To the best of our knowledge this is the first time NN-based control is being applied to design a controller for a WEC. NNs are a mathematical tool suitable to model the behaviour of dynamic systems, both linear and non-linear (as in our case). Significant absorbed wave energy increases were found, both using linear models and NNs. Results were better when IMC with NNs was employed (with a nearly sixfold increase against a fivefold increase), except for the May—September period, when IMC with linear models performs better.
- Multi-agent toolbox for fault tolerant networked control systems designPublication . Costa, José Sá da; Santos, Bruno M. S.; Mendes, Mário J. G. C.The design of fault tolerant control (FTC) systems of large-scale complex networked control systems (NCS) is a difficult task due to the large number of sensors and actuators spatially distributed and networked connected. Despite the research effort on developing FTC systems most of these developments are designed globally leading to centralized FTC solutions inadequate to NCS. In this paper we present the first version of a toolbox based on multi-agent systems (MAS) to design FTC systems for complex NCS. This toolbox is based on a decentralized FTC of NCS which relies on causal graph partitioning of the NCS digraph model and on intelligent distributed computing using MAS.