Grease Component Identification And Performance Optimization Based On Few-shot Learning

Lubricating grease is mainly composed of base oil,thickener and additives.Its composition and proportion determine the performance of lubricating grease.Traditional lubricant research mainly seeks the optimal formulation system through a large number of experiments,but this method is extremely time-consuming,labor-intensive and has high economic costs,which seriously restricts the cycle of grease research and development.With the development of information science,machine learning technology has spread to various fields.Its powerful data processing and analysis capabilities,combined with the relevant theory of small sample learning,can calculate the friction test data under laboratory conditions,which can meet the requirements of efficient,economic and predictable research on lubricating grease,so as to reduce the economic and time costs,and improve the design and production efficiency.The main work of this paper is as follows:(1)By changing the composite base oil ratio of mineral oil,hydrocarbon-based synthetic oil and synthetic ester,aiming at the change of the sample’s kinematic viscosity,viscosity index and rotary oxygen bomb value performance indicators,the performance prediction model based on regression algorithms such as GBDT is constructed and compared with the data enhancement as pretreatment;Various classical intelligent search algorithms are selected to optimize the parameters of the basic model,and the SMA-GBDT scheme designed has the best prediction results for the four performance indicators.(2)The three kinds of additives in grease were identified qualitatively,and the recognition model of grease additives was established by feature screening of full-band infrared spectrum data of samples.The infrared feature selection method(E-GA/S1-SVM and CARS)was used to optimize the features,and the combinatorial optimization model was established,the validity of the E-GA/Si-SVM-CRAS combination optimization model is proved by the validation of three kinds of commercial grease additives.(3)The extreme pressure composite lithium grease containing three additives(MoDTC,T321,T306)was prepared by orthogonal experimental design method.Through friction and wear experiments and calculation analysis,the performance prediction model was established and the additive formula was optimized.Firstly,the appropriate data features are selected by constructing polynomials;Then a regression model between additive content and tribological performance is established.PSO and MOPSO are used to optimize the objective function according to the constraint conditions,and the minimum friction coefficient and wear scar width are used as the target solution,and finally verified.