计算机 英文文献 文献翻译 基于ARM的系统的高精度温度测量的开发.docx

英文文献翻译 基于ARM的系统的高精度温度测量的开发 摘要 - 高精度温度测量系统，采用ARM处理器LM3S1138，作为控制的核心是本文介绍。在温度测量电路，Pt100A传感器的A / D转换电路的电流驱动电路采用相同的参考电压，这使得A / D转换结果只依赖上的铂电阻与温度的变化“RT的变化值，并传感器恒定电流驱动电流和A / D转换参考电压的精度的稳定性不会受到影响，设计方法，有效地提高了系统硬件的采样精度。在测量误差的软件校正，最小二乘法，以适应“采样值 - 温度”曲线在不同温度段的测量数据，偏移信号的各种中间环节发生的偏差，从铂电阻Pt100A温度测量A / D转换。在温度测量过程中，ARM处理器的实时计算，得到最终的温度测量结果，根据A / D的采样值，并结合小组拟合的多项式。实际测量结果表明，系统的温度偏差小于0.05。关键词：ARM，高精度的测量，温度检测电路，最小二乘法，曲线拟合。 一.导言温度是在工业生产中非常重要的参数之一，温度检测和控制精度，生产状况和产品质量的影响。因此，在许多工业领域，他们有温度测量和控制精度高的要求。在合并后的碱法化学工业的生产技术，外冷却器酒类进口和出口之间的温差是需要在一个固定值，低于或高于固定值，将有显著的负面影响。显然，较高的温度测量精度可保证温差控制精度。鉴于这种情况，该系统采用ARM处理器作为控制核心，采用铂电阻温度元件，温度信号的检测。虽然PT100有很多优点，如高精度，高稳定性等，在温度测量系统作为检测元件PT100，其测量精度是需要进一步改善，有存在一些实际困难。首先，在硬件电路方面，除了能够检测和放大信号的细微变化，它也必须能够补偿精度误差造成的，如零调整电路电路偏置电路等和温度漂移电路元件的工作期间，在同一时间，它有一个很高的要求在稍后阶段为铂电阻信号调理电路。另一方面，在温度测量系统的设计，一定规格的高温部件，一般都采取作为一个对象，但在铂电阻温度测量元件的批量生产，产品规格相同的错误是在一定范围内控制（如A级铂电阻Pt100A电阻可容许偏差是0.06在0），在同一规格的各种检测元件的差异被忽视，这使得成为不确定的允许偏差范围的温度测量系统的测试结果标称的检测元件，从而温度测量精度的提高是有限的。在上述基础上，推出后的温度测量系统的电路结构，硬件的信号检测电路，可满足需要高精密测量，然后使用一个特定的铂电阻，测得的数据，结合曲线拟合分析方法，就可以得到DT温度测量的多项式（D是A / D采样值，T是温度）这是适用于每个温度段的。 DT作为参数拟合方法可以抵消各种中间环节，从铂电阻Pt100A温度测量的A / D转换的信号发生偏差，使系统的温度测量更准确。 二系统硬件设计A. 系统硬件理论图表温度测量系统采用LM3S1138作为控制核心。 LM3S1138是基于ARM的Cortex - M3内核的第一个处理器。该处理器具有32位数据处理能力，其内部集成的可编程系统定时器（SysTick的），系统时钟可以乘以频率高达200MHz，这是足以满足在线（实时）计算需要的温度误差软件补偿。同时，它包含了丰富的片上外设：8个通用I/ O端口（GPIO）PORTA的PORTG，三个异步通信端口（UART0UART2），2个I2C总线接口（I2C0，I2C1的），32 - 位定时器和A / D转换器等，带来极大的方便连接其它外部电路。温度测量系统的硬件结构图如图1所示。信号放大单元，滤波电路，A / D转换和恒流源驱动电路，传感器的温度测量部分组成，采用液晶显示模块的嵌入式ST7920的信息显示控制器（基于LM3S1138模拟SPI的I / O端口软件与液晶显示模块的总线连接）;使用独立式键盘来设置系统参数;RS - 232转换电路与PC机通信，通过Modbus协议交换数据，从而可以实现PC的温度测量系统的监测。图1.温度测量系统的硬件结构图。B. 温度测量电路和分析温度测量电路主要由恒流源电路，信号放大，有源滤波和A / D转换电路，具体电路如图2所示。双运算放大器构成恒流源电路，如图2所示，电流通过RREF，即VREF / RREF，Pt100A驱动器的电流是： I= VREF/ RREF （1）图2.温度检测电路。精密运算放大器OP77构成一个信号放大单元2，以消除铂热电阻的接线电阻R的影响，Pt100A采用三线连接。取R0的100.00（也就是说，0 Pt100A在阻力），放大电路是电压信号，这是对应的“实时”，RT的电阻随温度变化时，铂电阻RT0变化作为一个参考点.当取R2= R3 R0输出电压后放大是：其中，K为放大器的增益。MAX7403是高阶低通有源滤波器，信号增益是在传输频段0分贝，频率高于截止频率衰减为- 80dB通过相应的信号增益。随着温度是相对缓慢变化的信号，设置过滤器38Hz左右，可有效过滤掉50Hz的电源线的干扰，造成系统电源的截止频率。MAX1169是16位串行A / D转换。它采用I2C总线与ARM处理器LM3S1138，LM3S1138是用来控制A / D转换和读取数据连接。由于通带内的预滤波电路对信号增益没有衰减，A / D器件的输入电压Vi是功放单元的输出电压Uo。由于A / D器件也作为参考电压VREF（恒流源电路），A / D转换结果D和ADC的输入模拟信号UO之间的关系如下所示：其中，n表示A / D转换位。提供的表达（2），（3）和（4）：显然，当放大单元的增益K和电阻值RREF是不变的，A / D转换结果只在不断变化的价值“的铂电阻与温度的变化RT依赖，传感器恒定电流驱动器的稳定性不会受到影响当前和A / D转换的参考电压的精度。 Pt100A传感器和A / D转换电路的电流驱动电路采用了一个参考电压VREF3，改善了电阻变化“实时有效地检测精度。 三.“样本值温度”曲线拟合和温度测量计算 高等教育的“RT检测精度可以得到式（4）。然而，在实际应用中，温度测量精度不够理想不能得到表达（4），有很多原因，如R和T之间的关系，并非是线性的，错误的运算放大器调零电路和偏置电路，滤波器的弱衰减，温度漂移和非线性的A / D等4。为了抵消在曲线拟合过程中的偏差，这是各种中间环节的A / D Pt100A温度测量转换造成的，之间的采样值这是A / D转换和温度T的关系获得，也就是说，铂电阻，后者作为一个整体来弥补类和A / D转换电路的信号调理电路。由于考虑到信号调理电路的错误，这DT曲线拟合的方法可以得到较高的温度测量精度比装修铂电阻R（t）曲线5。获得DT曲线后，如果可以读取采样结果，温度值可以通过在线计算。A. 采集和原始数据的预处理硬件设计，以了解这是用来适应“样本值 - 温度”的原始数据曲线。它以恒温水箱，温度场做十个连续抽样调查，在相同的温度点TI S1S10的抽检结果，并作为一组数据记录的抽检结果，S1S10。可以做数字滤波器处理后的原始数据进行了测量，测得的数据，排除在每个采样点的10个数据的最高和最低值，然后在这个数据的平均值其余8个数据作为最后的采样值DI点，因此，获得一组数据（DI，TI）。在实验中（约0.1在采样点），超过980套096节中的数据进行测量。因为它涉及到更多的数据点，为了便于表达，选择一个短节来形容。部分在2838日的实测数据见表1。B.实现曲线拟合 近似真实曲线的温度测量电路，方法分得到的拟合曲线应用在整个温度测量范围，即是用多项式，以适应在约10每个。多项式拟合具有以下形式6：其中，A0，A1，A2. .我将获得的系数。为了实现更准确的表达，三次多项式，以适应在数据处理。使用最小二乘法7，以确定多项式系数A0，A1，A2，A3拟合曲线。 DI数据测量结果的价值是非常大的表1，如果这是基本数据的DI是用于拟合，这将减少的准确性，因此，正常化8装修前完成DI的数据，是归表达：然后做装修后正常化的温度和样品的价值，获得拟合的多项式。 如果min（D）=13944，最大（D）=18371，在2838一节，后归和温度TI DI“之间的拟合函数是：在096截面积，在各个部分的多项式系数，获得配件如表2所示。表2.不同温度下的拟合多项式系数 096的部分C.在线计算的温度测量结果 LM3S1138 ARM处理器具有32位计算的准确性和足够高的工作频率，这是足以满足在线计算精度的温度测量结果和速度的要求。 DI是由采样值 A / D转换，选择相应的多项式系数，测得的温度进行实时计算。例如，如果采样值在13944，18371间隔（采样间隔对应2838温度节），精确的温度 测量对象的价值可以得到由式（6）及（7）。同样，其他温度段的精度测量结果可以通过采样值DI其中A / D转换。 四.测量ACCURARY及误差分析在不同温度段总的拟合曲线的误差平方如表3所示。为了更直观地观察错误消息，得到的拟合函数的A / D采样，收集原始数据采集阶段，在每个采样点的值被替换，计算相应的拟合温度值，然后有所作为宣读了在采样点的温度计和它本身的温度值之间，得到各温度点的偏差。采取各种温度点为横坐标，各点的误差，协调，误差曲线不同温度段的图表绘制。较小的错误部分和更大的错误部分的误差曲线图表分别显示在图3和4。其中，图3是较小的误差曲线在2838，图4是较大的误差曲线在4858。在图3和图4，“*”代表各计算点的测量误差。两行数字，得出的最大正负误差限制。从这些数字可以看到错误消息。最大误差在50附近出现，但不超过0.05。显然，安装后的测量值和实际测量值的温度计0.05之间的偏差。图3.在2838的误差曲线。图4.在4858的误差曲线。五.结论Pt100A传感器的A / D转换电路的电流驱动电路采用相同的参考电压Vref，保证系统硬件的温度的高精度检测;最小二乘法用于适合DT曲线，它的偏移量的偏差发生在各种中间环节Pt100A温度测量信号放大调理和A / D转换和其他中间环节，如，结合多项式拟合，使温度测量样本值，温度值，可在网上计算系统实现更高的精确度。参考1松井邦彦，实用的电路设计和制造传感器，北京：科学出版社，2005。2他洗菜和薛荣毅，传感器及其应用实例，北京：中国机械工业出版社，20043潘文成公主，王金亮，沉才中，王淑琴。 30MK精密铂电阻温度测量系统的发展。 2007年第八届国际电子测量与Instruments.p1613 - 1617（2007）会议。4 O. Kochan，R. Kochan，O. Bojko，M. Chyrkaand，基于控制温度场热电偶温度测量系统。 2007年第四IEEE研讨会onIntelligent数据采集和先进的计算系统：技术和应用，P 47-50（2007年）。5杨溶洲，“线性补偿，精度高，在智能仪器仪表及其实现铂电阻温度计”。仪表技术与传感器，8号，pp.44 - 46，2000。6陈Zhenglim，徐华“的调查督导工具TemperatureCompensation方法的研究”。杂志电子测量与仪器。第14卷，第3期，pp.9 - 12，2000。7尚宇培，石linsuo，张政仁，“应用最小二乘法温度测量精度高”。中国传感器技术，第19卷，第1期，pp.47 - 48，2000。8约翰Hmathews和芬克第纳尔，用数值方法MATLAB.4th版，北京：电子工业出版社，2005年。Development of High-precision Temperature Measurement System Based on ARMYi Xianjun, Liu Cuimei School of Electrical and Information Engineering, Wuhan Institute of Technology693 Xiongchu Avenue, Wuhan, 430073 P. R. ChinaEmail: Abstract The high-precision temperature measurement system which adopts ARM processor LM3S1138 as controlling core was introduced in this paper. In the temperature measuring circuits, the current driving circuit of sensor Pt100A and A/D converting circuit adopt the same reference voltage, which makes A/D conversion results are only dependent on the changing value of platinum resistance with the temperature change RT , and will not be affected by the stability of sensor constant current drive current and the reference voltage accuracy of A/D converter, the design method has improved the sampling accuracy of the system hardware effectively. In the software calibration of measurement error, Least Squares Method was used to fit the sampling value - temperature curve in connection with the measured data of various temperature sections, it offsets the deviation which occurred in various intermediate links of the signal from platinum resistance to the A/D conversion in the temperature measurement of Pt100A. In the temperature measurement process, ARM processor do real-time computing to get final temperature measurement results according to the A / D sample value and combined with the polynomial of Sub-fitting. Actual measurement results show that, the system temperature deviation is less than 0.05 .Keywords ARM, high-precision measurement, temperature detection circuit, least squares method, curve fitting. I. INTRODUCTION Temperature is one of very important parameters in the industrial production, the accuracy of temperature detection and control has impact on production status and products quality. Therefore, in many industrial fields, they have high demands for the accuracy of temperature measurement and control. In the combined-soda process production technology of the chemical industry, the temperature difference between the import and export of outer cooler liquors is required at a fixed value, lower or higher than the fixed value will have a significant negative impact. Obviously, the higher temperature measurement accuracy can guarantee temperature difference control accuracy. In view of this situation, the system adopts ARM processor as controlling core, uses Platinum resistance as temperature components, does detection for temperature signals. Although the Pt100 has many advantages, such as high precision, high stability etc, in the temperature measurement system which takes Pt100 as detection components, if its measurement accuracy is needed to improve further more, there are some practical difficulties exist. First, in regard to the hardware circuit, in addition to be able to detect and amplify the slight changes of signals, it also must be able to compensate precision error which was caused by circuits such as zero-adjusting circuit, bias circuit Etc and temperature drift during the work of circuit components, at the same time, it has a high requirement for signal conditioning circuit of platinum resistance at a later stage. On the other hand, in the design of temperature measurement system, Temperature components of certain specifications were generally taken as an object, but in the Volume production of Platinum resistance temperature measurement components, Products Error of the same specifications is controlled within a certain range(Such as A-grade platinum resistance Pt100A Resistance Admissible deviation is 0.06 at 0), the differences of various detection components at the same specifications was neglected, which makes test results of temperature measurement system become uncertain in the allowable deviation scope of nominal detection component, thus the increase of temperature measurement accuracy was limited. Based on the above, after the circuit structure of temperature measurement system was introduced, a hardware-signal detection circuit which can meet the needs of high-precision measurement was given, and then for a specific platinum resistance, use the measured data, combine with curve fitting analysis method, can get d-t temperature measurement polynomial ( d is A/D sample value, t is temperature) which was applied to the each temperature section. Its fitting method which takes d & t as parameters can offset the deviation which occurred in various intermediate links of the signal from platinum resistance to the A/D conversion in the temperature measurement of Pt100A, makes the systems temperature measurement more accurate.II. SYSTEM HARDWARE DESIGNA. System hardware theory chart Temperature measurement system adopts LM3S1138 as controlling core. LM3S1138 is the first processor based on ARM Cortex-M3 kernel. The processor has a 32bit data-processing capacity, its interior integrates programmable system timer(SysTick), system clock can be frequency multiplying up to 200MHz,which is sufficient to meet the online (real time) computing needs of the temperature error software compensation. At the same time, it contains a wealth of on-chip peripherals: eight general I/O ports (GPIO) PORTAPORTG, three asynchronous communications ports (UART0UART2), two I2C bus interfaces(I2C0,I2C1), 32-bit timer and A/D converter, etc, which bring great convenience to connect other external circuitry. The hardware structure chart of temperature measurement system as shown in Fig.1. Temperature measurement part is composed by signal amplification unit, filter circuit, A/D conversion and constant current source drive circuit based on sensor; adopt the LCD module of embedded ST7920 controller as information display (the I/O port software based on LM3S1138 simulate SPI bus to connect with the LCD module); use independent type keyboard to set system parameters;RS-232 converter circuits communicate with PC to exchange data through the Modbus protocol, which can achieve the monitoring of temperature measurement system by PC.Fig. 1. The hardware structure chart of temperature measurement system.B. Temperature measurement circuit and analysis Temperature measurement circuit is mainly composed of the constant current source circuit, signal amplification, active filtering and A/D conversion circuit, specific circuit as shown in Fig.2. Dual operational amplifier constitutes a constant current source circuit, as shown in Fig.2, the current through Rref is Vref / Rref , that is, the drive current of Pt100A is: I=Vref / Rref (1)Fig. 2. Temperature detection circuit.Precision operational amplifier OP77 constitutes a signal amplification unit2, in order to eliminate the effects of platinum thermal resistance wiring resistance r , Pt100A adopts three-wire connection. Take R0 100.00: (That is, the resistance of Pt100A at 0 ), amplification of the circuit is voltage signal which was corresponding to RT , RT is resistance change with temperature changes when platinum resistance RT take 0Cas a reference point. When take R 2=R3R0 output voltage after amplified is: In which, K is the amplifier gain. MAX7403 is high-order low-pass active filter, signal gain is 0dB in transmission bands, the signal gain corresponding to the frequency which was higher than the cut-off frequency attenuated by -80dB. As the temperature is the signal with relatively slow changes, set the cut-off frequency of filter around 38Hz, which can effectively filter out the 50Hz power-line interference caused by system power supply. The MAX1169 is 16-bit serial A/D converter. it uses I2C bus to connect with ARM processors LM3S1138, LM3S1138 is used to control A/D conversion and read data. As the signal gain of the pre-filter circuit within the pass band didnt attenuate, A/D devices input voltage Vi is amplifier units output voltage Uo . As the A/D device also take the Vref (of constant current source circuit) as the reference voltage, the relationship between A/D conversion result d and ADCs input analog signal Uo is shown as follows:In which, n represents the bits of A/D converter. Available by the expression (2),(3) and(4): Obviously, when the amplification unit gain K and resistance value Rref are constant, A/D conversion results are only dependent on the changing value RT of platinum resistance with the temperature change, and will not be affected by the stability of sensor constant current drive current and the reference voltage accuracy of A/D converter. The current drive circuit of sensor Pt100A and A/D converting circuit adopt a reference voltage Vref 3, which has improved the accuracy of the detection for resistance change RT effectively.III. SAMPLE VALUES -TEMPERATURE CURVE FITTING AND THE TEMPERATURE MEASUREMENT CALCULATIONHigher detection accuracy of RT can be get by expression (4). However, in the practical application, enough ideal temperature measurement accuracy can not be get by expression (4), there are many reasons, such as the relationship between R and t is not linear, the error of operational amplifier zero-adjusting circuit and bias circuit, filters weak attenuation, temperature drift and non-linear of A/D, etc4. In order to offset the deviation which was caused in various intermediate links of the A/D conversion in the temperature measurement of Pt100A, in the curve fitting process, the relationship between the sampling values d which was get by A/D conversion and temperature t was obtained, that is, take platinum resistance, signal conditioning circuit of latter class and A/D conversion circuit as a whole to compensate. Because of considering the error of signal conditioning circuit, this method of fitting d-t curve can get higher temperature measurement accuracy than fitting platinum resistance R (t ) curve5. After obtained d-t curve, if the results of sampling d can be read, temperature value can be obtained by online calculation.A. Acquisition and preprocessing of original data The hardware is designed to acquaint original data which was used to fit Sample values - temperature curve. It takes thermostatic tank as temperature field, do ten continuous samplings at the same temperature point ti to get sampling results S1S10, and take the sampling results S1S10 as a group of data records. After original data was measured, can do the digital filter processing to the measured data, exclude the maximum and minimum values of the ten data at each sampling points, and then take average value of the remaining eight data as final sample value di at this data points, thus, a group of data ( di ,ti ) was obtained. More than 980 sets of data in 096 section were measured in the experiment (About 0.1 at a sampling point). Because it involves more data points, In order to facilitate the expression, select one short section to describe. Parts of the measured data in 28 38 are shown in Table 1.B. The realization of curve fitting In order to approximate the true curve of temperature measurement circuit, the method of sub-obtained fitting curves was applied in the entire temperature measurement range, that is, use a polynomial to fit at each about 10. The polynomial fitted has the following form6: In which, a0, a1, a2am is coefficient which will be obtained. In order to achieve an more accurate expression, cubic polynomial was used to fit in data processing. Use Least square method7 to determine the polynomial coefficient a0, a1, a2, a3 of fitting curve. Data value of measurement results di is very large in Table 1, if di which is taken as basic data is used to fit, which will reduced the accuracy, therefore, normalization8 was done to di data before fitting, the normalized expression is: And then do fitting of the temperature and sample value after normalized to obtain fitting polynomial. If min( d ) =13944, max(d )= 18371, in 28 38 section, fitting function between di after normalized and temperature ti is: In 0 96 section area, the polynomial coefficients in various section were obtained by fitting as shown in Table 2. Table 2. Fitting polynomial coefficients of various temperatures section in 0 96 .B. Online calculation of temperature measurement results ARM processor LM3S1138 has 32-bit computing accuracy and a sufficiently high operating frequency, which is sufficient to meet requirements of the online calculation accuracy for temperature measurement results and speed. Sampling values di were get by theA/D converter, select the corresponding polynomial coefficients, the measured temperature was calculated in real-time. For example, if sampled values were at 13944, 18371