【作者】 李喜武；

【导师】 苏建；

【作者基本信息】 吉林大学 ， 载运工具运用工程， 2012， 博士

【摘要】 随着经济的发展和技术的进步，汽车的拥有量正日益迅速扩大，人们在享受汽车带来便利的同时，也在加剧不可再生能源的消耗。如何减少汽车能源的消耗是人们面临的迫在眉睫的课题。其中，保证汽车发动机能够工作在正常的润滑状态，就是减少能源消耗最好的办法之一。为了保证汽车发动机工作在正常的润滑状态，关键是必须保证在用润滑油的品质。最近，随着基于多传感器技术的信息融合技术的发展，已经在各个领域得到广泛的应用。与传统的单一传感器的检测技术相比，运用多传感器数据融合技术在解决系统分析判断、目标的跟踪、识别和探测等方面，具有提高系统的信息利用率和实时性、扩大系统在空间和时间的覆盖率、增强数据的精度和可信度、提高系统的鲁棒性和可靠性等优点。通过多个传感器获得的多层次、多方面、多级别的数据，经过处理后所能表达出的信息，比单一传感器获得信息更具有接近真实的意义，可作为各种系统决策的依据。信息融合技术应用到润滑油的监测之中，既能克服传统检测慢的缺点，又能克服单一传感器在线检测准确性差的缺点。本文以此理论为依据，先探讨了润滑油四项理化性能指标与其品质之间的关系，以及对其介电常数影响；其次探讨了红外光谱下润滑油的污染状况与品质之间的关系；再次探讨了润滑油的磁导率与其中铁磨粒含量的关系，然后探讨了超声波在润滑油中传播与其中铁磨粒含量的关系；最后综合以上各方面的因素，利用多传感器的信息融合理论，对润滑油的品质进行分析评价。首先，对润滑油各项指标与其介电常数的关系进行了研究。润滑油的理化性能指标包括水分、铁含量、酸值、不溶物含量等，在实验室配制以上四种指标的不同浓度的润滑油试验样本，考察不同浓度时所对应的介电常数，找到四种指标与介电常数的对应关系。确定以介电常数为评价润滑油品质的指标的可行性。其次，根据润滑油在使用过程中发生降解，其化学组成会随之发生变化，即氧化后形成的含有酸、酮、醛、醇等含氧有机化合物，以及含氮硝化物等官能团的量发生改变，通过红外光谱分析可知润滑油的降解程度。实质是依据烃分子对红外光谱具有吸收强度呈现某种可加性、特征谱带的光谱吸收系数近于常数，以及不同分子中相同结构的特征吸收峰几乎在相同的光谱区的特性，来测得碳原子在中芳环、环烷环及烷基链上的分配的。此外它还可检测油中某些添加剂和污染物含量。在实际试验中，监控的是光谱功率分布下的域，所有波长相关的变化，综合来评价润滑油的劣化程度。再次，润滑油在使用过程中铁磨粒的含量会逐渐增加，磨粒的含量、尺寸、集合形貌等因素，均会对润滑油的品质产生影响。因此利用磁导率测量方法测量磨粒的浓度。同时，由于超声波在润滑油中传播过程中，与其中的颗粒相遇时，一部分会射到颗粒的内部被吸收，另一部分会在界面散射衰减，而且在接触界面的超声波还会发生粘滞衰减。这些衰减的发生均是由润滑油中的铁磨粒引起的，而且与铁磨粒的数目即浓度成比例。因此利用超声波法测量润滑油的铁磨粒浓度。最后，进行数据融合分析。将在测量分析得到的润滑油的介电常数、红外光的透射值和散射值，以及利用磁导率法和超声波法测得的铁磨粒含量值，进行数据融合分析。分别建立两个子神经网络，对测得的数据进行处理。这样划分降低了每个神经网络的复杂程度，减小了诊断空间的维数，也降低了训练时间。然后进行D-S证据推理，将两个独立的低维的神经网络作为证据理论的一个证据，将其输出值处理后，作为辨识框架上命题的基本可信度，进行再次的融合。这样处理充分利用了信息源的信息，提高了润滑油污染度的判别精度，消除了单一数据源包含信息不全面的缺点。更多还原

【Abstract】 With the development of economy and the progress of technology, the number ofcar ownership is increasingly expanded. People are enjoying the convenience byautomobile and also aggravate the consumption of non-renewable energy resource.The pressing issue people face is how to reduce the consumption of automobileenergy. Ensuring the engine work under normal lubrication condition is one of thebest methods reducing the energy consumption. In order to guarantee the normal workof the automobile engine, the key is that the quality of the lubricants must be ensured.Recently information fusion technique based on multi-sensor technology hasbeen widely applied in all fields. Compared with the measurement technique based onsingle-sensor, the multi-sensor data fusion technique may improve the utilization rateand real-time of information system in solving system analysis judgment, the targettracking, identification and detection and so on. It may enlarge the coverage in spaceand time, advance the precision and reliability of system, enhance the robust andreliability of system, etc. The multilevel and aspects data obtained by multi-sensor ismore close to the true meaning compared with single sensor, can be also used as thebasis for the decision-making of the system.Information fusion technology applied to the monitoring oil can not onlyovercome the disadvantages of low testing speed and also improve the measurementaccuracy of on-line for single sensor. This paper based on this theory, the relationship between the quality and performance indicators of oil and the influence on dielectricconstant of oil are studied in paper.The relationship of structure group composition and quality of oil under theinfrared spectrum is also discussed. In finally the lubricating oil quality is analyzedand evaluated by the theory of multi-sensor data fusion and above two factors.First, the relationship between various index and dielectric constant of oil wasresearched. The index of physical and chemical characteristics for oil includemoisture content, iron content, acid value and insoluble matter content, etc. The testspecimens of various concentrations for above four indexes were mixed in lab. Thedielectric constant corresponding to various concentrations was examined andinvestigated the corresponding relationship between four indexes and dielectricconstant of oil. The feasibility to evaluate the oil quality index by the dielectricconstant was finally determined.Secondly, the chemical compositions of oil change with the degradation in usingprocess, the quantity of oxygen organic compounds contained of the acid, chromone,aldehydes and alcohols and functional group of nitrogenous nitration also change afterthe oil was be oxidized. The degree of degradation of oil may be showed by infraredspectroscopic analysis. The Essence is the distribution of carbon atoms in aromaticring, cycloalkane ring and alkyl chain. The basic is that some additivity of hydrocarbonmolecules may be showed to the absorption strength of infrared spectrum, spectralabsorption factor during characteristic spectral bands is nearly constant and theproperty that characteristic absorption peak with same structure in different moleculesis almost same spectral area. Furthermore, the method may also measure the contentof some additives and contamination in oil.The structure group composition, viscosity number and the value of moisturecontent was measured to show the indexes change in actual test.Finally, multi-sensor data fusion was analyzed. The acid value, moisture content,iron content and insoluble matter content obtained by measurement, and the viscosity,index of moisture content by infrared spectroscopic analysis was discussed by data fusion analysis. Two sub-neural network was established, a neural network to dealwith the physical and chemical indexes, the other to process the data by infraredspectra analysis. The complexity, dimension number and training time of NN wasreduced and by two NN. Then D-S evidential reasoning was carried out, two lowdimensions NN acted one of evidence theory. The output value processed wasregarded as the propositional basic reliability of frame of discernment and fused again.The precision of oil contamination degree and the utilization ratio of information resourcewere increased by this process. The faults that sole data source include imperfect information wasalso eliminated by it.更多还原