Paulo Afonso

Sensor Fault Detection and Identification in a Pilot Plant Under Process Control

The experimental evaluation of an automatic procedure for sensor fault detection and identification in a real process under closed-loop control is the objective of the present research. The scheme proposed here is very robust to faults in the main sensors of a multiloop control system, thus improving safety and reliability of plant operations. A state variable transformation is carried out in order to derive a model suitable for Recursive Least Squares (RLS) identification valid for all regimes of operation. The fault detection method is based on a moving window statistical analysis of the estimated model parameters. Simultaneously, a state estimation scheme, based on the Extended Kalman Filter (EKF), enables the fault identification, reduces false alarms and provides redundant measurements for alternative control purposes. Experimental runs were carried out in an industrial-scale pilot plant. Despite the large number of uncertainties and nonlinearities in the process, the system exhibited a good performance when faults occurred in the sensors of the control loops.

Model predictive control of a pilot plant reactor with a simulated exothermic reaction

Abstract The main goal of this work is to investigate the performance of a Receding Horizon Model Predictive Control scheme applied to a real plant. Nonlinear Model Predictive Control (MPC) algorithms have been described in the literature with great success when applied in simulation studies. However, simulation results do not always reflect the process realities, and they are often inadequate to guarantee that such sophisticated control algorithms can be applied in real plants. In this work, a linearized MPC control design has been experimentally used for temperature and level control in a CSTR. The process is unstable in open-loop, the variables are not all measured, while the process parameters are not very well known and can change with time. The experimental control tests were carried out in an industrial scale pilot plant where a pseudo zero order exothermic chemical reaction is partially simulated. The rate of heat generated by reaction is calculated and converted into an equivalent steam flowrate, which is injected in the mixture. A significant improvement in the controller performance is obtained with the MPC strategy, when compared with the previously existing PI based controller structure. The possibility of implementing the MPC in the corresponding industrial environment is outlined.

Prediction of Solar Radiation Using Artificial Neural Networks

Solar radiation data is needed by engineers, architects and scientists in the framework of studies on photovoltaic or thermal solar systems. A stochastic model for simulating global solar radiation is useful in reliable power systems calculations. The main objective of this paper is to present an algorithm to predict hourly solar radiation in the short/medium term, combining information about cloud coverage level and historical solar radiation registers, which increased the performance and the accuracy of the forecasting model. The use of Artificial Neural Networks (ANN) model is an efficient method to forecast solar radiation during cloudy days by one day ahead. The results of three statistical indicators - Mean Bias Error (MBE), Root Mean Square Error (RMSE), and t-statistic (TS) - performed with estimated and observed data, validate the good performance accuracy of the proposed three indicators.

Fall body detection algorithm based on tri-accelerometer sensors

In this paper a fall body detection system for a smartphone device is proposed. Its embedded tri-accelerometer sensor was utilized to collect the information about the body motion used by a real-time Pose Body Model (PBM) identified by an Extended Kalman filter algorithm. The PBM supply an estimate about the vertical pose angle value and a neural network is used to detect body fall incidents. Moreover, an automatic Multimedia Messaging Service (MMS) will be sent to a central of vigilant where additional information including the time and the GPS coordinates, reports the suspected fall location.

Body Fall Detection with Kalman Filter and SVM

In this paper, an approach for human body fall detection is presented that can be supported with a modern smartphone equipped with accelerator sensors. Falling is one of the most significant causes of injury, mainly for elderly citizens, and is one of the reasons why many individuals are forced to leave the comfort and privacy of their homes and live in an assisted-care environment. The acceleration measured by the embedded tri-accelerometer sensor, was utilized to collect the information about the body motion and was used to develop a robust algorithm to accurately detect a fall. This data is incorporated by a realtime Pose Body Model (PBM) which is identified by an Extended Kalman Filter (EKF) algorithm. Moreover, a Support Vector Machine (SVM) performs a binary classification of the observed data, allowing the detection of fall incidents. This fall detection system is tailored for mobile phones and has an important application in the field of safety and security, but can also be used in motion analysis of body moving and live style monitoring. Experimental results showed that this methodology can detect most types of single human falls quite accurately.

Improving safety of a pilot reactor using a model based fault detection and identification scheme

Abstract This work describes the experimental implementation of an automatic scheme for the on-line detection and identification (FDI) of faults in the sensors of an industrial scale pilot plant reactor under process control, where a pseudo zero-order exothermic chemical reaction is partially simulated. The main goals of this research are to enhance the safety of reactor operations and to demonstrate the potential of FDI for practical industrial applications. The automatic fault detection and identification method proposed here has two main steps: (1) the detection stage, which relies on a sequential statistical analysis of the process parameters that are continuously estimated by means of a general regression software package (GREG) suitable for non-linear models; (2) the identification step, which is based on an Extended Kalman Filter (EKF) to provide values for the state variables estimates. These values are compared to those given by the sensors thus enabling the identification of the faulty sensor. Moreover this classification procedure ensures that automatic process control can still be carried on even in such a faulty situation. Despite the strong non-linearities and the high number of uncertainties, the proposed strategy exhibited very promising results concerning the detection and identification of the faulty sensors. Furthermore, it enabled a satisfactory controller performance for a reasonable period of time, when any of the sensors was disabled and control actions were solely based on state estimates.

Separation of linguistic information methodology in a fuzzy model

A new methodology for organising the information in different structures was developed: the Separation of Linguistic Information Methodology (SLIM). The concept of relevance, which is in its genesis, was recently proposed, enabling the measurement of the relative importance of rule sets. Based on this methodology, a new algorithm was proposed for the Parallel Collaborative Structure (PCS). As demonstrated in the experimental tests, this new approach may benefit from the capability of recognised fuzzy systems to deal with natural language and model uncertainties. In addition, the proposed SLIM-PCS algorithm has been successfully applied to the modelling of a pilot plant reactor.

Solar Pyramidal Sensor

The use of solar radiation, as a renewable energy source, is a concern and a technological challenge of management and economic profitability. In this context it is paramount the knowledge about the position and movement of the Sun as well as the characteristics of sun radiation. Any new solution is required to bring a higher yield, better forecasting methods and better energy management, with lower technology costs and higher management effectiveness. This paper presents a simple and low-cost approach to determine the elevation and azimuth of the Sun, and the direct and diffuse radiation. This task is performed by a pyramid made of LDR sensor-coated surfaces connected to a microcontroller and a built-in mathematical model to compute the data. This device reads the radiation values of the sensors and sends them to a computer for further analysis, processing and presentation of useful information about the incoming sun radiation.

Hybrid fuzzy clustering neural networks to wind power generation forecasting

Wind power forecasting methods can be used to plan unit commitment, scheduling and dispatch by system operators and electricity traders. Because wind power is weather dependent, and therefore, is variable and intermittent over various time-scales, an accurate forecasting of wind power is recognized as a major contribution for a reliable large-scale wind power integration taking profit of economics gains. This paper explores a new approach using fuzzy clustering algorithms for obtaining one day forecast for the characteristics curves of speed wind. Moreover, a Feedforward Neural Networks (FNN) provides an estimate of the average hourly wind speed, for 24 hours horizon.

Decomposition of a Greenhouse Fuzzy Model

This paper describes the identification of greenhouse climate processes with multiple fuzzy models by resulting of decomposition of one global (flat) fuzzy model. This process is called separation of linguistic information methodology - SLIM. In this paper, the SLIM methodology is based on fuzzy clustering of fuzzy rules algorithm (FCFRA), which is a generalization of the well-known fuzzy c-means. It allows the automatic organization of the sets of fuzzy IF ... THEN rules of one fuzzy system into a multimodel hierarchical structure, result of clustering process of fuzzy rules. This technique is used to organize the fuzzy greenhouse climate model into a new structure more interpretable, as in the case of the physical model. This new methodology was tested to split the inside greenhouse air temperature and humidity flat fuzzy models into fuzzy sub-models.