PITKANEN, Jaakko1, VALJAKKA, Jukka2 & VIERINEN, Kari3
1 Espoo-Vantaa Institute of Technology, Finland, Jaakko.Pitkanen@evitech.fi
2 Jukka.Valjakka@evitech.fi
3 Kari.Vierinen@evitech.fi, http://www.evitech.fi/~karisv
Abstract: Espoo - Vantaa Institute of Technology (EVITech) is actively developing new approaches in the methods as well as in the contents of engineering education. This representation describes some of the learning projects related to sensor applications and imaging technologies developed and carried out in our institute.
The development of technology has provided new means to solve real-life problems also in a classroom setting. Modern and efficient computational tools for advanced signal processing, and use of high technology equipment to carry out measurements, have brought totally new prospects in education. We will respond to new challenges by taking into account working life requirements in a better and a more appropriate way. New thoughts of learning, discovery based learning, learning by doing, problem and project based learning, etc., shall bring business-like working and thinking also to part of learning process which in turn will lead to better qualified students, and at the end to better qualified engineers.
Our basic idea has been to approach learning process in a more problem oriented way contrary to the way how different disciplines used to consider problems only from their point of view. Real-life engineering problems are more or less interdisciplinary so we have taken into account aspects from planning experiments and modelling the process, via doing measurements and refining the obtained data into final analysis and assessments for the consistency of the model.
The present projects deal with modelling, use of modern high technology sensors in data acquisition, and processing the data by use of data refining methods (data compression, feature extraction etc.), and analysing the data by appropriate methods (discrete Fourier transform etc.). One of the tasks has been to classify the data into categories i.e. to infer the classes of objects involved by measuring accelerations and some other physical variables. This is basically a pattern-recognition problem.
On one hand modern high technology sensors together with computer interface give totally new possibilities to perform experimental measurements. The use of new sensor technologies in physics laboratory shall increase motivation, and link physics’ studies into practical applications. Many features of modern industrial measurements and control systems can be learned in basic physics laboratory. On the other hand modern computing tools like MATLAB, LABVIEW and MAPLE V make it possible, to use the techniques of scientific computing to solve realistic nontrivial problems even in a classroom environment, and learn advanced mathematical methods at a level adapted for engineering students.
Sensors
In EVITech physics laboratory we have used capacitive acceleration sensors, strain gage, Hall effect and piezo crystal force sensors, capacitive humidity sensors, many types of temperature sensors and many others. Microelectromechanical (MEMS) sensors are today available for standard physics laboratories. Infrared radiation have been studied with modern IR-camera [3]. For more detailed information on sensors see [12].
Infrared (IR) camera is based on modern platinum silicide (PtSi) focal plane array (FPA) detector operating at cryogenic temperatures (77 K) [3]. FPA technology requires no mechanical scanning. Instead, a mosaic of 65,536 discrete platinum silicide detectors, arranged in a pattern of 256 by 256 elements, is used. The PtSi -detector works like a photodiode, that is when infrared light of proper wavelength (2.0 m m to 6.0 mm) strikes each individual detector, the detector will change its resistance to a bias current imposed on it, and allow an increase in current flow. Current is proportional to infrared radiation energy striking the detector. More on IR cameras see [13].
Computer interfaces
Many interfaces between the sensors and computers have been used. Universal Laboratory Interface (ULI) [5] and Pasco interfaces [6] are used by all of our engineering students. For more advanced student projects we have used LABVIEW [7] , Nokeval [8], and some other interfaces and software tools. The two mentioned interfaces use 12-bit ADC’s so that several sensors can be connected at the same time to make real-time observations possible. The advanced interfaces which we have used have been 15-bit and 16-bit ADC’s with 16 to 96 possible sensor connections. We have used only 5-6 sensors connected simultaneously. The latest student project learning system is MAWS Automatic Weather Station which is now under construction by a group of engineering students [9].
We have had students projects where Visual Basic (v.6.0) has been used to build real-time applications for different sensors and interfaces. Some student projects have dealt with real-time www-applications [10]. For next semester we have plans to continue with sensor projects by using Visual Basic for database and www-applications.
Mathematical tools and applications
MATLAB software has become popular in all engineering fields so that, today, it can be considered as the world standard for the simulation and analysis of linear and non-linear dynamic systems, and as the most versatile numeric analysis toolbox. MATLAB along with other mathematical computing tools make it possible to use the techniques of scientific computing to solve realistic nontrivial problems even in a classroom setting. These problems have been traditionally avoided, since the amount of work required to obtain a solution exceeded the classroom time available and the capabilities of the students. This situation has changed and students can be taught with real-life problems that can be solved by powerful software tools. This has created us new possibilities to learn mathematics by using computers.
In some student projects we have considered applications where the measured variables have been slowly varying and also corrupted by random noise. It has been desirable to apply several smoothing filters to remove noise from the signal, and reconstruct the underlying smooth function for further analysis. The methods used at this stage have been quite simple for pedagogical reasons so that students could have ruled the different phases of the problem and got an insight into engineering problem solving. Students have had to face quite a challenge to find proper means to deal with the measured data which as such has not been ready for further analysis.
Students have applied Discrete Fourier Transform (DFT) –methods to implement filters (e.g. low-pass) into MATLAB (acceleration (elevator, respiration) and humidity measurements). They have also implemented floating mean filter to smooth humidity data, which has been successful in the applications. Students have also encountered some basic problems with sensors which have not been stable during the measurements but have showed some roaming tendency so that they have had to model the phenomenon by using low order polynomial fitting curve to compensate the roaming error.
Image Enhancement Project
Blurred images give real challenges to an engineer to improve the quality of the image to some approved standards. We have made quite an effort to find out basic principles on image enhancement, and also to simulate the degrading process of an image.
In our project we simulated the degrading process of an image by adding white noise on it. To remove noise by Fourier technique is based on the fact that essential information of the image is concentrated on lower frequencies than those of the noise. However, in our examples this method did not work. Because of the large pixel size image information was gathered in the same frequencies than the noise.
Improving resolution by deconvolution
Our image model was based on convolution principles
The function E represents the effects of the lens system of the camera on the target function O defined on a plane, and I is the image function created by the camera. Our goal is to solve the target function O provided that we know the kernel E and that we have got the image. We have made some assumptions on the camera that it exposes a limited region so that regions outside the image-area do not affect on the image by any profound quantity so that O will vanish outside a restricted area. By applying the convolution theorem we can write down the convolved version of the above integral equation
from which we can in principle solve for the target function by inverse Fourier transform
In practise this formula is totally useless because of its sensitiveness to disturbance.
Typically function E converges towards zero with large values of 
. When I containing measurement errors is divided by values close to zero, the measurement errors are magnified in such an amount that the image becomes but mere noise. We have found a regulated pseudo solution for the problem.
Let’s suppose that we know the magnitude of the noise level of the image in the sense of 
-norm so that 
. In principle by using the information available, any solution O which satisfies the above inequation, is as good a solution as the original O. We will find out the solution which minimises -norm, and satisfies the inequation.
By a fixed value a > 0 the pseudo solution 
should satisfy the equation
from which we can find an immediate solution 
. We will at once notice the basis for regularisation by realising that the denominator is always positive. By substituting this solution into equation 
we will get a simple condition 
. The last stage is to choose a value for a
which corresponds to the noise level. We will do this by iterative Newton’s method by finding zero for the function k.
Elevator
In this project study, different sensor measurements were completed with elevators in a testing facility builded by Kone Corporation [11]. The elevators move a vertical distance of about 310 meters below the ground level, and this offers very special conditions for many different sensor applications. The measurements were carried out by a common elevator model and a top speed high-tech elevator. The dynamics and vibrations of the elevator motion were studied by using accelerometers along with measurements of pressure, humidity, magnetic field etc.
Students have analysed the altitude of the elevator by measuring the pressure inside the elevator by a barometer and by establishing a simple mathematical model for pressure with due corrections
The analysis explains well the known facts of the elevator system (maximum speed, the range of the elevator etc.).
Figure 1. The barometer sensor measurement. The distance of the motion of the elevator calculated from the pressure variation. The constant speed of the elevator is very nicely demonstrated.
Loops with an aeroplane
Acceleration sensors have been used in many study projects and experiments. Students have studied acceleration of elevators, local trains and cars. Also collision experiments, jumping, free fall and many other experiments have been made. One of our latests acceleration experiments were done by Cessna 150 aeroplane in loops. We used two Pasco 500 data collecting and storing units [6] and two 3-dimensional acceleration sensors [5]. The measured resultant acceleration is presented in figure 2. Two complete loops were measured and one half loop, and a turnaround at the highest point. The interval for the first loop was 49 - 80 s and for the second one 82.5 - 103 s.
Figure 2. Resultant of the acceleration of the Cessna 150 airplane.
The half loop started at time 104.5 s. One of the authors of this article Kari Vierinen was inside the airplane.
IR-camera
IR -cameras have been used to analyse temperature leaks in buildings, temperatures of electric wires and cables, temperatures in electronic circuits, heat flow in differents systems, condition of mechanical bearings etc. With the IR-camera images can be taken on videotape to study dynamical effects stored on PC digital storage cards and they can be later analysed with computer software tools from the database of digital image information. In Figure 3 is the image of our Physics laboratory fuse panel. It can be seen that some of the fuses are rather warm and the phases are not loaded evenly.
Figure 3. The IR image of the Physics laboratory fuse panel.
We have learned quite a lot about new things in engineering applications, and we think so have our students as well. However, we have still long way to go to improve the learning process. Problems have to be defined more specifically in student projects. Project based learning is not always very efficient for all students. Engineering applications and projects from the real life have opened new possibilities and have brought more motivation for students in their Physics and Mathematics studies.