322
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO. 1, FEBRUARY 2005
Self-Organizing Maps Application in a Remote Water Quality Monitoring System Octavian Adrian Postolache, Member, IEEE, Pedro M. B. Silva Girão, Senior Member, IEEE, J. Miguel Dias Pereira, Member, IEEE, and Helena Maria Geirinhas Ramos, Member, IEEE
Abstract—This work was developed in the context of a system for remote water quality monitoring based on a wireless local area network (WLAN) and includes a Kohonen self-organizing map (K-SOM) implementation in order to perform sensor data validation and reconstruction and sensor failure and pollution event detections. Simulation and experimental results are presented. Index Terms—Self-organizing maps, telemetry, water quality monitor.
I. INTRODUCTION
S
EA AND RIVER water quality monitoring is one of the important activities in the environment-monitoring domain. The number of research and development activities in that area is extremely large. The assessment of water river basin conditions for drinking water is reported by the Guilikeng investigation group [1]. Two networks based on supervision control and data acquisition (SCADA) systems were used to collect data from automatic monitoring stations located on riverbanks. Meanwhile, different commercially available monitoring systems such as a remote underwater sampling station (RUSS) [2] or YSI [3] are used in different water quality monitoring applications in order to collect multiple water quality parameters (pH, temperature, conductivity, turbidity, heavy metal concentration, etc.) from rivers, lakes, or sea water. The acquired data is usually sent to a central location using mobile phone [global system for mobile communication (GSM), global packet radio service (GPRS) [4], or personal handy-phone system (PHS) [5])], satellite, or VHF [6] technologies. Water quality monitoring hardware (sensors, conditioning circuits, acquisition, and communication) must usually be complemented with processing blocks (WQ-PB) to perform different tasks associated to one-dimensional or multidimensional data that flows on the system measuring channels. Important processing tasks of the WQ-PB are data validation, data linearization [7] and data compensation [8], short and long term prediction of pollution events (duration and concentration), data fusion and data compression, fault and pollution detection, and data recovery. One of the most successful solutions for advanced WQ data processing is neural networks. Thus, different applications of neural network on
Manuscript received June 15, 2003; revised May 29, 2004. This work was supported by Portuguese Science and Technology Foundation PRAXIS XXI Program FCT/BPD/2203/99 and by Project FCT PNAT/1999/EEI/ 15052. The authors are with the Instituto de Telecomunicações, Centro de Electrotecnia Teórica e Medidas Eléctricas, DEEC—Instituto Superior Técnico, 1049-001 Lisboa, Portugal (e-mail:
[email protected]). Digital Object Identifier 10.1109/TIM.2004.834583
water treatment [9] and on river water for drinking purpose [1], are reported in the literature. This paper deals with the design and implementation of a distributed measuring system for water quality monitoring and emphasizes multilayer perceptron and Kohonen maps neural network processing structures implemented for advanced data processing. The system is characterized by high accuracy of water parameter measurements, data validation, data reconstruction, sensor fault detection, and pollution events signaling. Data flow obtained from the primary acquisition and processing units is distributed in a wireless local area network (WLAN). The WLAN includes a personal computer (PC) as a central processing unit that manages the distributed measuring system and analyzes the received data. II. WQ MONITORING SYSTEM DESCRIPTION The remote distributed measuring system includes a set of FieldPoint primary acquisition and processing units (PAPs). Each PAP contains a processing unit with an ethernet controller interface (FP-2000) and a four–channel analog input module (FP-TB-10) that performs the water quality parameter acquisition. Communication between the PAPs and the central control and processing unit (CCP) is based on WLAN and GPRS [10] technologies. Thus, the FieldPoint units that include the RS232 and ethernet interfaces can be directly connected to the CCP in order to transmit the numerical values of water quality parameters from the zones to monitor. In the following two subsections, two network and communication architectures are presented: a GPRS polling architecture (GPRS-PA), which was initially selected, and a WLAN-GPRS hybrid architecture that is the upgrade of GPRS-PA finally implemented. A. GPRS Smart Polling Architecture The GPRS polling architecture includes a set of Siemens Cellular Engines M35 [11] that are class B GPRS mobile stations. The M35’s only port (RS232) is used to connect it to the FP2000 of the PAPs (Fig. 1). Data acquired by the PAPs is sent to the CCP using GPRS M35 terminals and is then processed by the CCP’s personal computer. The CCP interrogates the PAPs from time to time according to the derivative values of the measured quantities in order to obtain more accurate data for fault and pollution events detection (smart polling). The instantaneous sampling rate of a water quality parameter has a nominal value of for . However, the a stationary signal
0018-9456/$20.00 © 2005 IEEE
POSTOLACHE et al.: SELF-ORGANIZING MAPS APPLICATION IN A REMOTE WATER QUALITY MONITORING SYSTEM
Finally, the sampling frequency equation
323
can be obtained using the
with
(3)
Equation (3) ensures that the computational and communication load of the system is very low under normal working conditions but increases with the time rate of the measured quantity. Howmust be established in ever, a maximum sampling interval order to detect global communication failures and to assure that a minimum amount of data is gathered for WQ parameters time mapping definition. B. WLAN-GPRS Hybrid Architecture
Fig. 1. Water quality monitoring architecture based on FieldPoint embedded systems and GPRS modems.
Fig. 2. Water quality parameter linear variation.
sampling rate is proportionally incremented by the relation between the absolute value of the water quality parameter variaand its nominal value . The factor , where tion is the maximum sampling interval , works as a speed-up coefficient for the sampling rate when large variations of water quality parameters occur. In fact, this speed-up that depends on system effect is limited by the value of hardware and system configuration, namely, the number of PAP ms. units. In our practical implementation, Considering a linear approximation for the water quality parameter as the one represented in Fig. 2, the absolute value of the straight line slope is equal to (1) The normalized slope used to increment the sampling rate in dynamic conditions is given by (2)
In order to reduce costs, to be able to increase the distance between the CCP and the monitored area and, at the same time, to increase the communication rate up to 11 Mbps between field units should communications between the PAPs and the CCP fail, a hybrid architecture based on a WLAN and on GPRS modems was designed and implemented. Data acquisition rates are slow, but increasing the communication rate capability among PAP units can be very important to detect system faults, to identify geographical and temporal trends associated to pollution events, and also to increase system protection against faults (robustness) if GPRS communication failures between the access point (AP) and the CCP unit occur. In this case, one of the PAP units assures the management and the control of the measuring system and temporarily performs the functions associated with the CCP unit. The communication hardware is WLAN-based and includes a set of 2.4-GHz ethernet-to-wireless bridges (D-LinkAir DWL810) that convert the FP2000 ethernet port into an IEEE 802.11b wireless network device. Because wireless outdoor transmission range is lower than several hundred meters (e.g., 300 m), additional wireless bridges can be used to extend the range, availability, and functionality of the implemented WQ wireless network. For the same reason, and in order to increase the dism), two GPRS tance between CCP and PAPs (e.g., range modems (Siemens M35) are used: one connected to the CCP and the other to the PPP Gateway. The WLAN-GPRS compatibility is assured by the D-Link DWL-900AP wireless access point, by the ipEther232.PPP PPP-Gateway, and by a M35 GPRS modem (Fig. 3). The PAP information is sent wirelessly to the access point (AP) that communicates via GPRS with the CCP. The AP protocol is compatible with IEEE 802.11b, and its operating frequency and modulation are 2.4 GHz and Direct Sequence Spread Spectrum (DSSS), respectively. The PAP works as a bridge between the FieldPoint WLAN and the wired PPP Gateway (ipEther232.PPP). The gateway enables the transmission of network packets via the serial interface of the GPRS modem. The serial port of the PPP Gateway accommodates baud rates from 2400 to 115 200 baud and is connected to an M35 GPRS modem that assures the communication with the CCP. The GPRS modem enables a transparent connection between the CCP and the PAPs that are identified by their IP addresses.
324
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO. 1, FEBRUARY 2005
Fig. 3. Upgrade of the architecture of Fig. 1. Water quality monitoring system based on a WLAN-GPRS hybrid architecture.
C. Primary Acquisition and Processing Units Each PAP incorporates several hardware blocks: i) sensing block; ii) acquisition block; iii) primary processing and communication control unit; iv) communication block. The sensing block includes a set of sensors and conditioning circuits associated to water quality (WQ) parameters measurement such as turbidity (TU), pH, temperature (T), and conductivity (C). The WQ values are transmitted as a current amplitude (4–20 mA) to PAPs analog inputs. The sensors used are Global Water WQ770 for turbidity measurement, ISI OLS50 for conductivity measurement, ISI-11 for pH measurement, and a Pt1000 for temperature measurement. The acquisition block incorporates two FieldPoint dual channel analog input modules FP-AI-V10B (12 bits, up to 2 kS/s) mounted on a FP-TB-10 terminal base and a set of 4–20 mA receivers RCV420 that convert the current signals , , and . into voltages The acquired data is delivered via the FieldPoint bus to a primary processing and communication control unit (FP2000) that performs embedded measurement, data logging, and communication tasks. The acquired data is stored in nonvolatile memory for reliable data logging or is transmitted to the CCP, as described before, using the serial RS232 interface (first architecture) or the ethernet communication interface (second architecture) that are
part of the communication block. Other elements of this communication block are a GPRS modem, for the first considered monitoring system architecture, and a wireless bridge for the second one. D. Central Control and Processing Unit The central control and processing unit includes a PC (Pentium III, 128 Mb RAM) and several communication interface components. Thus, for the first proposed architecture a GPRS modem (Siemens M35) assures the direct communication between the PC and the GPRS compatible field units (M35 GPRS modem included). In the second architecture, the CCP communicates via GPRS with the networked PAPs using a PPP Gateway (ipEther232.PPP) and a M35 GPRS modem that perform the WLAN-GPRS conversion. The data flow, GPRS received by the CCP, is distributed over the Internet, the CCP working like a WQ Internet Server. Based on the GPRS connection, the CCP controls the WLAN elements (PAPs) in order to acquire new WQ values that permit the actualization of a WQ Net Page. Anomalous functioning of the PAP’s channels is detected by the CCP, which then issues commands to switch off the anomalous channels. III. SOFTWARE COMPONENTS OF WQ MONITORING SYSTEM PAPs and CCP software components were mainly developed using LabVIEW 6.1 and LabVIEW Real Time [12]. MATLAB 6.0 modules for offline MLP-NN design and optimization were implemented using the Neural Network toolbox [13] and the
POSTOLACHE et al.: SELF-ORGANIZING MAPS APPLICATION IN A REMOTE WATER QUALITY MONITORING SYSTEM
n
TABLE I MLP-NNS PARAMETERS: n —NUMBER OF INPUT NEURONS; -NUMBER OF HIDDEN LAYER NEURONS (ONE HIDDEN LAYER) AND —NUMBER OF OUTPUT NEURONS n
SOM toolbox [14] used for the K-SOM’s design and online operation. The connection between MATLAB and LabVIEW modules was realized using MATLAB Script LabVIEW structures [15] that permit the use of the MATLAB functions within LabVIEW programs. The tasks performed by the software are data acquisition and processing at PAPs’ level (including PAPs’ measuring channels modeling [16] and compensation of temperature effects [17]), advanced analysis of WQ data at the CCP level, data communication, and data publishing. A. PAPs’ Software The software associated with PAP units includes the acquisition subroutines, based on the FP Read LabVIEW—Real Time function, and neural processing blocks [multilayer perceptron neural networks, (MLP-NN)], designed using MATLAB, that , , , and acquired voltages into convert the values of TU, pH, C, and T, respectively. Thus, two single and input-single output neural networks (one for conversion) and two dual input-single another for temperature output neural networks [one for temperature comcompensated pH and another pensated C conversion] were designed and implemented. The networks have one single hidden layer, the training algorithm in all four cases is Levenberg Marquardt [18] and sum-square errors (SSE) of 1E-4 are the stop condition. The MLP-NN’s training and test are performed using a data set of voltage values delivered by the transducers during PAP’s calibration phase. The calibration solutions used were formazin NTU for the TU transfor the pH transducer, ducer, buffer solutions uS/cm for conductivity. Calibraand tions were performed in a laboratory for different temperatures, C. Taking into account the FP2000 storage memory limitation, an optimization of the MLP-NNs was carried out in order to obtain high conversion accuracy using a small number of neurons. Table I presents not only the results obtained in terms of the numbers of neurons for each of the three layer MLP-NNs but also of neurons transfer functions, tan-sigmoid (tansig ( )), for the hidden neurons, and linear, for the output neurons. After MLP-NNs design in MATLAB, the calculated neuron weights and biases are then used to implement the online neural conversion and compensation blocks based on the following relation: (4)
325
TABLE II PERCENTAGE ERROR OF WQ PARAMETERS VALUES FOR THE DATA RECEIVED FROM THREE PAPS AFTER NEURAL NETWORK PROCESSING. IN BOLD, AND FOR COMPARISON PURPOSES, ARE THE PERCENTAGE ERROR OF THE RAW WQ PARAMETER VALUES (WORST CASE ONLY)
where and represent the weights matrices, and and represent the biases matrices of the neural network designed using the acin order to obtain the WQ parameter value , , pH , and included in quired normalized voltages vector . The conversion and compensation blocks are parts of LabVIEW Real Time program uploaded to the PAP’s FP2000 memory. The global errors associated to WQ measuring channels are lower than 1% (Table II) in Tagus River WQ operational meaNTU conductivity surement ranges TU mS/cm temperature C and confirm that MLP-NNs do increase the accuracy of WQ evaluation, particularly when the measuring channels are highly nonlinear (e.g., conductivity measurements). The PAPs also include signaling blocks associated to pollution events and faulty operation detection. For that purpose, a comparison is made between the current TU, pH, C, and T measured values and the historical values (last 2 h) of the measured values and the WQ parameter limits for the monitored area (e.g., estuary of Tagus river) according to the European Environment Agency (EEA). Several LabVIEW real time blocks were designed and implemented at the FP2000 level for communication purposes. For the first architecture, the main software blocks are associated with RS232 port configuration and with sending and receiving data. The Serial Init LabVIEW function was used to set the serial communication port (e.g., 57.6 kbaud, one stop bit, non parity, no flow control), while the Serial Write function was used to program the GPRS modem using the AT commands according to the communication requirements. The Serial Read LabVIEW function was used to receive the commands issued by the CCP to control data flow and on/off switching of the measuring channels. For the second communication architecture, known as DataSocket, a programming technology based on industry-standard TCP/IP is used to perform data read and data write actions. Taking into account possible network failures, PAP unit software also includes a watchdog time function that autoreconfigures the network if faulty communication between the PAPs and the CCP are detected. As mentioned before, in this case, the connection with the CCP is disabled, and one of the PAPs assumes the management and control of the measuring system. After fault recovery, the CCP stops all PAPs programs, saves management data from “master” PAP, and reinitializes activities (Fig. 4). B. CCP Processing Component The CCP receives the WQ parameter values based on the PAP’s interrogation, processes the data, and publishes the WQ current values on the PAP’s web pages. Data processing is mainly
326
Fig. 4.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO. 1, FEBRUARY 2005
WQ network reconfiguration when CCP-PAPs connection fails.
related with multichannel data analysis and is based on Kohonen maps. The main objective of such processing is to reduce all WQ measured parameters to a format better suited to the extraction of information concerning water quality in the monitored area and the performance of the overall telemetric system. Kohonen maps are neural networks widely used in both data analysis and vector quantization because they compress the information while preserving the most important topological and metric relationship of the primary data and also for their abstraction capabilities. These characteristics of Kohonen maps are used in the present application for representation and analysis of WQ data received via WLAN—GPRS from the PAPs. (TU, Each K-SOM defines a mapping from the input space C, pH, T, and WQ parameters) on to a regular two-dimensional array of nodes. The number of designed K-SOMs is equal to the number of PAP units; three, in the present case. The K-SOMs were trained using the WQ parameters (TU, C, pH, and T) acquired by the PAPs. The used values correspond to the following conditions: “Normal Functioning (NF)” (the WQ parameters values are inside the EEA limits [19]; “Fault Event (FE)” (the WQ values corresponds to faulty functioning of one or several PAPs measuring channels); and “Pollution Event (PE)” (the WQ values correspond to either low values of pH or high water conductivity and turbidity values). Different were considered according numbers of cells to the size of the training set. While MLP-NNs used for data conversion and temperature compensation were trained using a supervised training strategy based on a gradient descent type algorithm, K-SOMs are obtained using an unsupervised learning process, which is an incremental-learning SOM algorithm in our application. Thus, the of cell of a K-SOM network randomly prototype vector initialized is updated according to the following learning rule: (5) represents an input vector randomly drawn from the where TU pH , and input data set at time , is called the neighborhood kernel around the winner cell and defined by (6) is the learning rate at time , is the distance between cells and within the output space (map), corresponds to the width of the neighborhood function and [20]. A batch-training algorithm was also applied, and a comparison in terms of resolution and topology preservation of the designed K-SOMs was carried out. The map training was implemented using the MATLAB SOM toolbox. As mentioned before, K-SOMs are used in the present application mainly for classification purposes: NF, FE, and PE. The degree of confidence of the classification depends on the comreceived from a PAP (inparation between the WQ vector cluding TU, pH, C, and T current values) and a reference vector using a Gaussian kernel defined by
Fig. 5. PAP1 web page including the TU, C, pH, and T information.
(7)
POSTOLACHE et al.: SELF-ORGANIZING MAPS APPLICATION IN A REMOTE WATER QUALITY MONITORING SYSTEM
327
Fig. 6. PAP’s K-SOM graphical representation.
The reference vector is the vector resulting from the training phase that describes the cell whose distance to the current input vector is a minimum. The corresponding cell is identified by is, the so-called best match unit (BMU). The closer to unit the higher the probability that the input vector is assigned to the , a weak association between the correct cluster. For input vector and the correspondent BMU must be assumed. to the distance Based on the contribution of , the faulty channel of a PAP and pollution events can be detected. In order to distinguish between faulty events and pollution events, a comparison between the values acquired at the same instant in different PAPs must be performed. C. WQ Data Publishing The water-quality data publishing is based on the LabVIEW web publishing tool. Values transmitted by the PAPs to the CPP are published on pap1.htm, pap2.htm, and pap3.htm pages. In Fig. 5, the PAP1 web page associated to TU, C, pH, and T data publishing is shown. Each PAP can directly generate its own web pages that can be visualized using an ordinary web browser (e.g., Internet Explorer). PAP’s web page generation offers the possibility to access online the evolution of water quality parameters using the PAP IP address (e.g., IP 193.136.143.199 for PAP1). However, the limited processing capability of the field units means that the PAP’s pages contain only information related with the current values of WQ parameters and, thus, do not offer additional information about global evolution of water quality based on Kohonen maps. Also, pollution events or anomalous functioning of PAPs channels are not analyzed at the PAPs level and, thus, are not presented on their web pages. Only the CCP with its
higher computation capability can perform advanced data analysis including data mapping, data recovering [21], and fault and pollution alarm generation. IV. RESULTS AND DISCUSSION The WQ information is transmitted to the CCP where the selforganizing maps are offline designed and implemented in order to perform the mapping of the WQ values received from the PAPs’ measuring channels in the operational phase. Based on training data associated to the PAP units (PAP1, PAP2, PAP3), a set of three K-SOMs was designed and is presented in Fig. 6. In that figure, the K-SOM cells correspond to the following situations: normal functioning (NF) of PAPs and normal water conditions (WQ inside EEA limits); fault events (FEs), related to anomalous operation of one or several PAP measuring channels (e.g., pH, C, T, or TU channel); and pollution events (PEs), related to WQ parameter values out of EEA specifications. In order to express the map quality dependence on the size of the training data, PAPs’ maps were designed using different numbers of WQ parameters measurements (4 624 for PAP1, 4 360 for PAP2, and 4 174 for PAP3). The number of map cells was experimentally optimized in order to obtain maps with the same number of cells but well-defined clusters, which lead to the 6 12 cells using MATLAB SOM Toolbox som_make( ) function. In Fig. 6, it is possible to observe the division of maps in clusters of cells: clusters that mainly corresponds to NF, FE, and PE situations and whose number of useful cells depends on the size of the training set (47 for PAP1, 34 for PAP2, and 32 for PAP3). Fig. 7 shows that the cluster borders are characterized by larger distances between neighboring cells, while the cluster
328
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO. 1, FEBRUARY 2005
TABLE III K-SOM1: EXAMPLE OF WATER PARAMETERS VALUES, CORRESPONDING BMUS, AND CONDITIONS
TABLE IV PAP’S K-SOM QUALITY EVALUATION
Fig. 7. PAPs D-MATRIX graphical representation: black
= 1 and white =0.
areas are characterized by low distance between neighboring cells. Distance D between a cell and its neighboring cells is calculated taking into consideration all the elements of the unified of a map. In order to obtain distance matrices , , and the D values, the MATLAB som_umat( ) function was used. Once the map designed, visualization of each WQ input vector implies the determination of its localization in the map. For that purpose, the so-called best-matching unit (BMU) must be evaluated. BMU is the value of that minimizes the distance
where takes all values between 1 and 72, in our case. If there is more than one with the same distance, then the winning cell is randomly chosen among them.
In order to express the correlation between PAPs operating condition and BMU values, a set of three WQ vectors associated with NF, FE, and PE conditions was used. For the particular case of K-SOM1, the correspondence between WQ parameter values and BMU is presented in Table III. The BMU value in time gives global information about the time evolution of WQ parameters acquired by the PAP units. The global information provided by the time evolution of the BMU value can be used to predict undesired events such as pollution events. Comparing the evolution of the BMU position in the maps, false pollution alarms can be reduced and, at the same time, the WQ evolution for a larger area can be evaluated. Transient alarm conditions are always very difficult to detect, especially when the working condition corresponds to a BMU that is in the border of different clusters (e.g., “PE” and “FE” zones). However, some success can be obtained in false alarm detection by analyzing the trajectory of the BMU in a given map or correlating the measurement data information from different maps. Normal system behaviors correspond to exchanges in a BMU position between the adjacent cells. Thus, when large jumps between BMU locations are detected, this is generally associated to a fault event condition. PAP’s K-SOMs quality evaluation criteria used in the present work were resolution and topology preservation. Thus, calcu(average distance lation of the average quantization error between each data vector and its BMU) and topographic error (percentage of data vectors for which the first- and secondBMUs are not adjacent cells) was carried out. The results are presented in Table IV and show that the iterative training algorithm is better because a) it leads to smaller values of the topographic error, which means a better separation between NF, PE and, FE clusters, and b) the training time and computational load required are lower.
POSTOLACHE et al.: SELF-ORGANIZING MAPS APPLICATION IN A REMOTE WATER QUALITY MONITORING SYSTEM
V. CONCLUSION Even if some interesting solutions in the domain of water quality monitoring are addressed—use of FieldPoint and Wireless LAN technology, WEB publishing capabilities using data socket transfer protocol implemented in LabVIEW, increase of the measuring channel accuracy using Multilayer Perceptron neural networks—the main contribution of this work is related with the utilization of Kohonen self-organizing maps (K-SOMs) for multidimensional data representation, data validation, and reconstruction. The results obtained are encouraging and point to the possibility of implementing accurate, standalone smart (self tested) water-quality monitoring systems with space and time integrating capabilities and, thus, of water-quality forecasting.
REFERENCES [1] A. M. Guilikeng, “Real-time monitoring and assessment of river basin conditions for drinking water intake protection,” in Proc. Water Quality Resource Management, REC-Telematics, Brussels, Belgium, 1999. [2] Remote Underwater Sampling Stations. Apprise Inc, Duluth, MN. [Online]. Available: www.apprisetech.com [3] YSI Environment—YSI-6600 EDS—YSI Catalog, Yellow Springs, OH. [Online]. Available: www.ysi.com [4] J. Hoffman, GPRS Demystified. New York: McGraw-Hill, 2002. [5] Public Personal Handy-Phone System, PHS MoU Group. General Description of Interworking/Internetworking of Public PHS Network, Tokyo, Japan. [Online]. Available: www.phsmou.or.jp/documents/pdf/B-IW0.00-03-TS.pdf [6] “Spread-Spectrum radio modem for fieldpoint,” in Meas. Automation Cat. Austin, TX: Nat. Instruments Inc., 2002, pp. 508–509. [7] J. Patra, A. Kot, and G. Panda, “An intelligent pressure sensor using neural networks,” IEEE Trans. Instrum. Meas., vol. 49, pp. 829–834, Aug. 2000. [8] O. Postolache, P. Girão, and M. Pereira, “Neural network in automatic measurement system: State of art and new research trends,” in Proc. IEEE—IJCNN, vol. 3–4, 2001, pp. 2310–2315. [9] N. Valentin, T. Denoeux, and F. Fotoohi. A Hybrid Neural Network Based System for Optimization of Coagulant Dosing in a Water Treatment Plant. Univ. Technol. Compiegne, Compiegne, France. [Online]. Available: www.hds.utc.fr/~tdenoeux/congres/icnn99.pdf [10] O. Postolache, P. Girão, M. Pereira, and H. Ramos. An internet and microcontroller-based remote operation multi-sensor system for water quality monitoring. presented at Proc. IEEE Sensors. [CD-ROM] 2002. [11] M35 GPRS Modem- Data Sheet. Siemens, Munich, Germany. [Online]. Available: www.siemens.com [12] J. Travis, LabVIEW for Everyone. Upper Saddle River, NJ: PrenticeHall, 2002. [13] (2003) Network Network Toolbox. Mathworks Inc., Natick, MA. [Online]. Available: www.mathworks.com [14] J. Vesanto, J. Himberg, E. Alhoniemi, and J. Parhankangas. (2000) SOM Toolbox for Matlab, Helsinki, Finland. [Online]. Available: www.cis.hut.fi/projects/somtoolbox [15] R. Bitter, T. Mohiuddin, and M. Nawrocki, LabVIEW: Advanced Programming Techniques. Boca Raton, FL: CRC, 2001. [16] J. C. Patra, A. Bos, and A. C. Kot, “An ANN-based smart capacitive pressure sensor in dynamic environment,” Sens. Actuators, vol. 86, pp. 26–38, 2000. [17] O. Postolache, M. Pereira, P. Girão, M. Cretu, and C. Fosalau, “Application of neural structures in water quality measurements,” in Proc. IMEKO World Congr., vol. IX, 2000, pp. 353–358. [18] S. Haykin, Neural Networks—A Comprehensive Foundation. Upper Saddle River, NJ: Prentice-Hall, 1999.
329
[19] Water Temperature Level of Pollution in Surface Waters for Quality Class I. European Environment Agency, Copenhagen, Denmark. [Online]. Available: http://www.eea.eu.int/main_html [20] T. Kohonen, Self Organizing Maps, 3rd ed. New York: SpringerVerlag, 1997. [21] T. Bohme, “Sensor failure and signal reconstruction using auto- associative neural networks,” in Proc. ICSC/FAC Symp. Neural Computation, 1998, pp. 220–225.
Octavian Adrian Postolache (M’99) was born in Piatra Neamt, Romania, on July 29, 1967. He received the electrical engineering degree from the Faculty of Electrical Engineering, Technical University of Iasi (TUI), Iasi, Romania, in 1992. In 1992, he joined the Faculty of Electrical Engineering Iasi, Department of Electrical Measurements,TUI, as an Assistant Professor, where he is currently Auxilliary Professor. In the last four years, he has developed research activity at Instituto Superior Técnico of Lisbon, Lisbon, Portugal and the Instituto de Telecomurncações, where is currently Senior Researcher. His main research interests concern intelligent sensors, laser systems, and intelligent processing in distributed measurement systems.
Pedro M. B. Silva Girão (M’00–SM’01) was born in Lisbon, Portugal, on February 27, 1952. He received the Ph.D. degree in electrical engineering from the Instituto Superior Técnico, Technical University of Lisbon (IST/UTL), in 1988. In 1975, he joined the Department of Electrical Engineering at IST/UTL, first as an Assistant Professor and, since 1988, as an Associate Professor. Presently, his main research interests are instrumentation, transducers, measurement techniques, and digital data processing. Metrology, quality, and electromagnetic compatibility are also areas of regular activity, mainly as auditor for the Portuguese Institute for Quality (IPQ), Lisbon.
J. Miguel Dias Pereira (M’02) received the degree in electrical engineering from the Instituto Superior Técnico (IST), Technical University of Lisbon (UTL), Lisbon, Portugal, in 1982. In 1995, he received the M.Sc. degree, and in 1999, the Ph.D. degree in electrical engineering and computer science from IST. He worked for eight years for Portugal Telecom, Lisbon, in digital switching and transmission systems. In 1992, he returned to teaching as Assistant Professor in Escola Superior de Tecnologia of Instituto Politécnico de Setübal, Setübal, Portugal, where he is, at present, a Coordinator Professor. His main research interests are in the instrumentation and measurements areas.
Helena Maria Geirinhas Ramos (M’03) was born in Lisbon, Portugal, in October, 1957. She received the M.Sc. and Ph.D. degrees in electrical engineering from the Instituto Superior Técnico of the Technical University of Lisbon (IST/UTL) in 1987 and 1995, respectively. In 1981, she joined the Department of Electrical Engineering at IST/UTL, first as Assistant and, since 1995, as a Professor. Her main research interests are in the area of instrumentation, transducers, and measurement techniques.
