温度校正的NaCl水溶液浓度超声检测装置设计与试验

Design and test of high accurately measuring equipment for NaC l water solution utilizing ultrasonic velocity with temperature correction

为了提高Na Cl水溶液浓度超声在线检测精度,该文设计了试验装置,利用该装置对Na Cl水溶液质量分数进行了测定。试验装置包括一对中心频率5 MHz的压电陶瓷超声换能器、电脉冲发射接收器CTS-80787PR、恒温水浴器、内壁距离5 mm的有机玻璃测试池和采样频率为40 MS/s的16位PCI总线数据采集卡,装置的声速测误差小于1 m/s,有效保证了Na Cl水溶液浓度测量精度。在10~30°C梯度为2.5°C的温度范围内以及0~10 g/100 g梯度为0.5 g/100 g的质量分数范围内,配置了标准Na Cl水溶液。开发LABVIEW应用程序,实现数据采集、分析、建模和验证。经残差分析去除异常样本点和t检验后得到质量分数的普通多元线性回归模型,模型的校正决定系数Rcal2达0.9992,校正标准差RMSEC（root mean square error of calibration）为0.0894 g/100 g,验证决定系数Rval2达0.9988,预测标准差RMSEP（root mean square error of validation）为0.1015 g/100 g,相对预测性能RPD（ratio performance deviation）为28.57,最大预测误差小于0.25 g/100 g,提高了检测精度。以上数据表明模型具有较高的预测精度和稳定性,为食盐水溶液质量分数的超声高精度在线检测提供参考。

In food, chemicals and pharmaceuticals industries, the concentration is a very important parameter for quality control. Continuous process monitoring is a fundamental requirement for the process control in these industries. Besides process parameters, such as temperature, pressure, liquid level, flow rate, concentration measurement is also of special interest. More qualified information obtained from new or better sensors can significantly enhance the process quality and thereby product properties. Low-intensity ultrasonic sensor system can contribute to this development, and it has been widely used in many fields for its characteristics are non-destructive, non-invasive and rapid. In this study, a method based on ultrasound for determining the concentration of NaCl water solution with high accuracy was presented. The ultrasound velocity in different temperatures and in solutions with different concentrations was measured using an emitter-receiver method. The experimental set-up was composed of two matched piezoelectric transducers with the center frequency of around 5 MHz （one used as emitter and the other as receiver）, a thermostatic tank, a CTS-80787PR pulse emitter-receiver （used to excite and receive electric signals） and a data acquisition card. The electric pulse at the emitting voltage of-25 V, the width of 100 ns and the repeated frequency of 100 Hz was sent from the CTS-80787PR to the emitter transducer and then transformed to ultrasound wave. This wave traveled through a liquid sample cell with 5 mm wide to reach the receiving transducer, and was transformed back to electric pulse signal. The signal was acquired by the CTS-80787PR and sent to a computer equipped with the LABVIEW via a 16-bit data acquisition card of PCI （peripheral component interconnect） bus with the sampling rate of 40 MS/s. A Labview interface was developed to collect signals, analyze signals, build models and validate models. Through analysis, the measurement accuracy of ultrasound velocity was 1 m/s for this equipment, which ensured the accuracy of concentration measurement. At 2.5°C gradient in the temperature range of 10-30°C, and at a gradient of 0.5 g/100 g in the concentration range of 0-10 g/100 g, NaCl water solutions were prepared. Some were regarded as calibration samples and others as validation samples. In order to predict the concentrations of NaCl water solution, a calibration model was established using a quadratic polynomial approach based on the experimental results. The coefficients of this polynomial were obtained based on ordinary multiple linear regression （OMLR） method by inserting the measurement points, analyzing the residual to eliminate abnormal sample points and checking out the regression coefficient of the model by t-test. The developed model had high prediction accuracy and stability with the maximum prediction error of 0.25 g/100 g, the determination coefficient of calibration （Rcal2） of 0.9992, the determination coefficient of validation （Rval2） of 0.9988, the root mean square error of calibration （RMSEC） of 0.0894 g/100 g, the root mean square error of prediction （RMSEP） of 0.1015 g/100 g and the ratio performance deviation （RPD） of 28.57, which indicated that the model could be used for practical detection accurately and steadily, and was helpful for on-line measuring.