Search

Sufficiently large amount of data is needed to prepare a training set for a neural network. However, there is no enough experimental data of thermocouple leg errors to prepare an adequate training set for the neural network. The method of preparation of a training set for the neural network is proposed in this paper. This method is based on hierarchical interpolation and extrapolation of the experimental data of thermocouple leg error using the method of conditional least squares. The neural network model of thermocouple leg error is presented on the example of the chromel leg of type K thermocouple.

Thermoelectric (TE) device has some problems, such as low mechanical strength, high manufacturing cost, etc. We are developing flexible TE device using thin films and flexible substrates in order to overcome those problems. The purpose of this work is to fabricate and evaluate the flexible TE device. A flexible TE device consists of vapor-deposited thin films of n- and p-type TE materials between two flexible substrates. The flexible substrate is made up of a combination of polyimide and copper. This structure enables the device itself to convert temperature difference of outer surfaces (DeltaTout) into in-plane temperature difference (DeltaT in), and it can generate electricity by the p-n couples. We analyzed temperature distribution of the device by the finite element method (FEM). According to the analysis, a flexible substrate which has thinner polyimide layer (h1) and thicker copper layer (h1) is preferable to obtain larger temperature difference. We fabricated flexible TE devices containing 33 pairs of p-n couple, which is composed of chromel and constantan thin films. The device of h 1/h2 = 12 mum/70 mum generated 3.72 muW at DeltaTout = 22.7 K, while 2.40 muW was generated from the device of h1/h2 = 35 mum/25 mum at DeltaT out = 24.0 K. The performance of the flexible TE device was successfully improved as predicted by the analysis

The purpose of this work is to fabricate and evaluate a flexible thermoelectric (TE) device. Because of its flexibility, it is expected that application field of the TE device will be enlarged. The device consists of p and n-type thin films placed between 2 polyimide sheets, and it can generate electricity from temperature difference (/spl Delta/T/sub out/) applied between the outer surfaces of the device. The /spl Delta/T/sub out/ is transformed into in-plane temperature difference (/spl Delta/T/sub in/), which can be utilized for thermoelectric power generation by the p-n couple. The /spl Delta/T/sub in/ was measured as 16.7 K when /spl Delta/T/sub out/ was 57.8 K, being in good agreement with calculated value by Finite Element Method. Calculations revealed that the use of TE material with high power factor is effective, and low thermal conductivity will be not so important for this device performance. Chromel as p-type and constantan as n-type was, therefore, selected for the flexible TE device. The device composed of 9 pairs of the films was fabricated by RF-sputtering-method. The device could generate maximum output of 144 nW at /spl Delta/T/sub out/ = 28.6 K in measurement. According to the calculation, maximum out of 114 nW at /spl Delta/T/sub out/ = 28.6 K was predicted, being in good agreement with measured value.

We report the fabrication and characterization of a novel type of metal-on-polymer micromachined thermopile. The thermopile is based on a design that consists of many three-dimensional mesas that are coupled in series. These mesas are produced by lithographically structured SU-8 deposited on silicon wafers. Identical copies were replicated into thermoplastic materials by injection moulding. By evaporative coating of Alumel and Chromel from two equal but opposite angles the two alloys will, due to shadowing effects, overlap each other only at the top of the mesas and between them. These two areas will therefore form the hot and cold solder junctions. Typical height and width of a mesa were 200 /spl mu/m and 300 /spl mu/m, respectively. The deposited metal was about 100 nm thick at the top, less metal would not form a conductive coating over the whole mesa structure. Scanning electron microscopy was used to investigate the surfaces of the micromachined mesas in SU-8 and thermoplastics, which in both cases showed very smooth surfaces. We measured the generated thermovoltage over many thermocouple elements coupled in series, and evaluated the voltage dependency of mesa height and number of mesas. These pilot tries demonstrate that our concept for a low-cost polymer-based thermopile element generates a measurable thermovoltage. More detailed experiments will be carried out in order to further characterize the polymer thermopile, especially with respect to sensitivity, noise equivalent power and response time, all three important figures of merit for a thermopile. Such a thermosensing element can be used in various applications where miniaturization is essential, for instance, infrared detection of handheld gas sensors, liquid flow sensors for microfluidics, or calorimetry for monitoring chemical or biological reactions.

A special chromel wire (C) and alumel wire (A) thermo-pile has been developed which can measure human respiratory condition and body temperature without directly contacting a sensor to the human body. The measurement system enables high speed, real time, noninvasive, and simultaneous measurement of respiratory rates and body temperature with the same sensor. The special CA thermo-pile with a sensing each junction of approximately 25 /spl mu/m were constructed by using spot welded thermo-pile junction. The thermo electric power of 17 pairs of special CA thermo-pile is 0.7 mV//spl deg/C. The special CA thermo-pile provides high sensitivity and fine frequency characteristics, of which the gain is flat to approximately 10 Hz. >