鲜食葡萄属浆果类，具有易腐性，在运输过程中容易出现腐烂、褐变、干枝、掉粒等问题，造成经济损失。冷链运输是解决此问题的有效途径。冷链运输过程中，冷藏车厢环境状态决定鲜食葡萄贮藏期和货架期的长短，如何准确、实时、低成本地监测冷藏车厢环境参数是保障鲜食葡萄冷链运输顺利进行的关键。因此，开展鲜食葡萄冷链运输过程的智能化动态监测方法研究，确保冷链运输过程中监测参数的有效采集、传输与处理，对推动我国鲜食葡萄冷链运输业的高速发展有着重要意义。基于此，本书以红地球葡萄为研究对象，以无线传感器网络技术、多传感器数据融合技术、灰色理论、神经网络理论等为数据获取和处理手段，结合保鲜技术、温度场分析方法等对鲜食葡萄冷链运输监测方法进行研究，主要结论如下:第一，温度、相对湿度和二氧化硫体积浓度是影响鲜食葡萄运输品质的主要因素，是冷链运输需要监测的参数。冷藏车厢温度的空间差异性、货物的装载方式、开关门时间长短等是引起这些参数发生变化的主要原因。第二，提出了多目标决策模糊物元法，实现了冷藏车厢内传感器布局的点位优化。结果表明，该方法能够在保证监测结果准确的前提下，使传感器数量从27个减少到7个，降低了冷链运输成本。采用统计分析方法和温度场分析法，验证了优化后传感器布局合理性。优化后传感器监测值具有95%以上的置信水平，优化前后温度场分布图相似率达到90%以上，达到冷链运输中既节约成本，又准确监测的双重要求。第三，提出了时间—空间数据融合算法，对冷链运输监测值进行融合估计，实现了冷藏车厢环境状态的准确反映。结果表明，时间—空间数据融合算法可以利用修正方差自适应调节各组传感器的修正因子，削弱较大误差对监测结果的影响，融合精度要优于分批估计法和算数平均法，处理后的结果能够更准确地反映冷藏车厢内的环境状态。第四，提出MGM－RBF神经网络预测模型，能够对冷藏车厢环境状态进行准确预测。结果表明，均方根相对误差为0.60%，平均相对误差为0.44%，明显优于单一的MGM预测和RBF神经网络预测。基于统计过程控制理论对冷藏车厢温度进行判异，定义了三种预警模式，对比固定阈值法能够明显减少误警或者频繁报警现象，提高了报警的准确度，实现了分类、分级、分层次报警目的。第五，从关键参数的辨识、参数的获取、参数的估计以及状态预报等方面对鲜食葡萄冷链运输过程展开研究，形成了一套高效实用的冷链运输动态监测方法。 关键词:监测　鲜食葡萄　冷链运输　数据融合　神经网络

The phenomenon of decay，browning and decay is easy to appear during the table grape transportation，which causes no little financial loss.Cold－chain transportation is an effective measure to prevent those problems.At present，the cold－chain transportation construction is still in initial stage in china.It is urgent to conduct a research on the intelligent monitoring technology of coldchain transportation to complete data collection and processing.The way guarantees the quality of the table grapes and promotes the rapid development of table grapes cold－chain transportation.In the paper，red globe grape was chosen as the research object.Wireless sensor network，multi－sensor data fusion technology and neural network theory were used for data acquisition and processing method combined with the preservation technology to guarantee the security of table grapes cold－chain transportation.By means of investigation research，emulation and simulation，some methods were put forward.Main conclusions were as follows:(1) Temperature，relative humidity，and volume concentration of sulfur dioxide are the main factors influencing the quality of table grapes transportation，and the three factors were chosen as the the main monitoring parameters in the cold－chain transportation.Spatial difference of temperature and operation nonstandard were the main reasons for those factors changes.(2) Multi－objective fuzzy matter element analysis method was put forward to optimize the sensor quantity.27sensors was reduced to 7 in the refrigerator car，which reduced the cold－chain transport costs.The statistical analysis method and the temperature field analysis were applied to validate the rationality of the optimization algorithm.Results possessed more than 95% of confidence level.(3) A new data fusion method based on time－space data fusion theory was investigated to multi－sensor data processing.The mentioned algorithm could gradually reduce the influence of sensors with poor precision by means of introducing correction factor into the weighting coefficient，which taked advantage of the variance of the single－sensor fusion data and the final fusion data to adaptively adjust the weights of each sensor，and by means of multi－step fusion，gradually weakened the influence of some sensors with larger errors on fusion accuracy.Experimental results showed that the proposed algorithm outperforms traditional method.(4) MGM－RBF neural network predictive model was put forward to predictive the refrigerator compartment temperature.Results showed that mean square relative error was 0.60% and average relative error was 0.44%，that were superior to single MGM arithmetic and RBF neural network prediction arithmetic.Based on the theory of statistical process control，three early－warning modes were defined.Compared with the fixed threshold method，the proposed algorithm had a less false－alarm rate.(5) Monitoring research involved four dimensions: key parameters identification，parameters acquisition，parameters estimation and parameters forecast.Those formed an efficient and practical method for cold－chain transportation dynamic monitoring. Keywords: Monitoring; Table Grapes; Cold－chain Transpotation; Data Fusion; Neural Network