外文翻译----多传感器的自我测量诊断系统

《外文翻译----多传感器的自我测量诊断系统》由会员分享，可在线阅读，更多相关《外文翻译----多传感器的自我测量诊断系统（8页珍藏版）》请在毕设资料网上搜索。

1、附录 A A Multi-Sensor Based TemDerature Measuring System with Self-Diagnosis Abstract- A new multi-sensor based temperature measuring system with self-diagnosis is developed to replace a conventional system that uses only a single sensor. Controlled by a 16-bit microprocessor, each sensor output from

2、the sensor array is compared with a randomly selected quantised reference voltage at a voltage comparator and the result is a binary “one” or “zero”. The number of “ones” and “zeroes” is counted and the temperature can be estimated using statistical estimation and successive approximation. A softwar

3、e diagnostic algorithm was developed to detect and isolate the faulty sensors that may be present in the sensor array and to recalibrate the system. Experimental results show that temperature measurements obtained are accurate with acceptable variances. With the self- diagnostic algorithm, the accur

4、acy of the system in the presence faulty sensors is significantly improved and a more robust measuring system is produced. Index Terms-Instrumentation and Measurement, Sensors. Transducers I. INTRODUCTION Conventional sensing system uses a single sensor to convert a measured into an electric signal.

5、 There is no built-in redundancy and the system is wholly dependent on the single sensor for its accuracy. Recently, a novel approach proposed by the author in l makes use of the principles of successive approximation and statistical estimation to provide a simple yet accurate estimate of the measur

6、ed with only a small number of sensors. Replacing the single sensor with a multi-sensor array also improves the robustness of the system reducing system dependency on any single sensor. The system is still functional even with a few faulty sensors, though there will be a degradation in the accuracy

7、of the results. To overcome the degradation in the accuracy due to the presence of faulty sensors, a self-diagnostic algorithm is devised to determine and isolate faulty sensors so that these sensors are not used in the determination of the temperature estimate. In this paper, the development of suc

8、h concept into a practical system for temperature measurement is described. II. SYSTEM ARCHITECTURE AND OPERATION A. System Hardware Architecture The hardware system consists of 36 temperature sensors in a ensure array, a signal conditioning circuit and a 16-bit micro- controller, as shown in Fig. 1

9、. Each sensor, controlled by an analog switch, measures temperature and outputs a voltage. The output from all 36 sensors are fed into a switching circuit. The switching circuit consists of a decoder and an analog multiplexer that is controlled by the software to sequentially select an output from a

10、ll the 36 sensors. The selected output is fed into the signal conditioning circuit for processing before being sent to the microcontroller. One complete “read cycle” involves reading the outputs from all 36 sensors. The sensors used in the sensor array are calibrated beforehand to obtain their volta

11、ge-temperature characteristic. The aggregate voltage-temperature relationship for the sensor array was found to be linear over the temperature range to be measured, thus a simple linear equation is used in the software algorithm to convert the voltage reading into a temperature reading. B. Temperatu

12、re measurement To obtain an estimate of the output temperature, mathematical principles of successive approximation and statistical estimation are used. The analog sensor output are sequentially selected by the switching circuit and passed onto the non-inverting input of a voltage comparator for dig

13、itization. A reference voltage that is determined by the software program is applied to the inverting input of the voltage comparator. If the analog sensor voltage is higher than the reference voltage then the output at the comparator is a binary “one”, else the result is a binary “zero”. The initia

14、l reference voltage range of is established based on apriority knowledge of the characteristics of the temperature sensors and the temperature range to be measured. The voltage range is then quantized into m different levels with an equal step sue of where m is the number of sensors in the sensor ar

15、ray and represents the maximum and minimum value of the initial voltage range before any successive approximation is carried out. The m reference voltages are randomly sorted. For each reading from the sensor array, a quantized reference voltage is randomly selected for comparison at the voltage com

16、parator. This is to reduce the dependency of any sensor reading to the reference voltage applied. At the output of the comparator, a binary “one” or “zero” is produced. The quantized reference voltage is generated by the software algorithm and converted into an analog voltage through a 12- bit digit

17、al-to-analog converter PAC. One complete “read cycle” involves processing the analog sensor voltages him all 36 sensors to obtain 36 binary readings. The binary output from the comparator is fed to the microcontroller for so hare processing. The microcontroller counts the number of binary “ones” in

18、a read cycle. Based on the number of “ones”, statistical estimation is used to obtain the temperature estimate V as follows. If the accuracy of the estimate, given by Arid, does not meet a predetermined level, successive approximation is carried out to reduce or narrow down the reference voltage ran

19、ge. The new reference voltage range for the new “read cycle” is given . where V is the current estimate of the sensor output, Avert ,o is the current quantized voltage step size and k is an integer that controls the next reference voltage range (in this case, k = 2). The new voltage range is again q

20、uantized into m levels with a voltage step sue of Avail and another “read cycle” is carried out to obtain a new estimate of the sensor output, The successive approximation is carried out until the required accuracy is obtained. In general, the reference voltage range, step size and the estimated sen

21、sor voltage output after the successive approximation (i = 1,2,.) are given by: The final voltage estimate is converted into a temperature reading using the conversion formula obtained from the initial calibration. Apart from calculating the temperature estimate, the software process is also respons

22、ible for synchronizing the various hardware components to ensure that the sensor readings are processed in the correct order. The flowchart in Fig. 2 shows the software algorithm for temperature measurement. Overall, the system works by digitizing the analog signals from each sensor in the array. St

23、atistical estimation is used to obtain a first approximation of the temperature. Successive approximation is then applied based on the estimate to reduce the voltage range until the desired accuracy is met. The process is repeated until the temperature measurement with a desired accuracy is obtained

24、. III. SELF-DIAGNOSIS The self-diagnosis algorithm is a software controlled procedure to detect whether any of the sensors in the array is faulty, to isolate and deactivate any faulty sensors present and to compensate for the faulty sensors. The diagnosis assumes that the majority of the sensors in

25、the array are in good order. A sensor is classified as faulty if its measurement is more than x from the actual temperature .where x is a user defined value depending on the temperature sensor used and the accuracy required. In the prototype. In diagnosis, the same reference voltage is applied to th

26、e voltage comparator for all the sensor outputs in a “read cycle. In principle, all the sensors are expected to produce the same digital out& with two exceptions: when the reference voltage is very close to the voltage corresponding to the actual temperature ； if the sensor is faulty and gives rise

27、to an inaccurate output different film those of the majority sensors. Thus as the reference voltage applied to the comparator is shifted between him V (min) to V (max), it should be able to separate the good and faulty sensors. This is illustrated in Fig. 3 where, as indicated, sensor 20 and 26 are

28、filly. The diagnostic algorithm works as follows. At a certain temperature, a constant reference voltage is applied to the comparator for all the sensors in the array. The total number of “ones” and “zeros” are counted and the majority state or “zero” is determined. A sensor that belongs to the mino

29、rity state is likely to be faulty. Each sensor has a software status counter associated with it that is initially set to zero at the start of the diagnostic routine. This counter is incremented if the sensor was found to belong to the minority state. Incrementing the status counter of a sensor indic

30、ates a high probability that the sensor is faulty. In Fig. 3 both sensor 20 and 26 will have their software counters incremented for the V, level since they belong to the minority state. The constant reference voltage is shifted between the extreme ends of the voltage range through scanning (moving from V(min) to V (max) in fixed