【作者】 唐昆；

【导师】 黄红武； 盛晓敏；

【作者基本信息】 湖南大学 ， 机械设计与理论， 2007， 硕士

【摘要】 钛合金以其比强度高,抗腐蚀性能优异,无磁性、耐热性能好、疲劳强度高等诸多优良特性,在各领域的应用越来越广泛。然而钛合金优良的材料性能,也给其磨削加工造成了很大困难。采用常规方法磨削钛合金,磨削力大、磨削温度高,砂轮粘附和损耗严重。因此,迫切需要采用新的高效高质量的先进磨削工艺对其进行加工。高速及超高速高效深磨作为一种先进的现代磨削技术,以超乎一般想象的极高磨削效率,极大的砂轮磨削比,良好的工件表面完整性,给传统的磨削领域带来了一场革命,展现了现代磨削技术发展的巨大潜力和广阔应用前景。将高效深磨技术应用到钛合金磨削加工中对于扩大高效深磨技术的应用范围,探寻钛合金高效磨削技术等方面都有重要意义。本文在40Cr钢高效深磨试验的基础上,对TC4钛合金进行了系统的高效深磨试验研究。试验中通过改变磨削用量,测量了两种材料在高效深磨条件下不同工况的磨削力和工件表面粗糙度。对比研究了它们的磨削力、比磨削能和磨削力比随磨削用量变化的特征,对它们反映出的材料去除机理和磨削力的形成机理进行了探讨,并推导了粘附条件下TC4钛合金高效深磨磨削力的数学模型;对比磨削能的特征、形成及磨削功率的消耗也进行了研究;分别采用传统BP神经网络和进化神经网络建立了工件表面粗糙度的预测模型,并对两者的预测结果作出了分析和比较。通过研究发现,高效深磨条件下,单位面积磨削力受磨削用量影响较大。单位面积磨削力、比磨削能与最大未变形切屑厚度、当量磨削层厚度之间存在一定的对应关系。两种材料的去除方式都以塑性去除为主,且在工作台速度和切深较大的情况下材料主要以滑擦和耕犁的形式被去除,两种材料因其本身性质的不同在磨削力、比磨削能上存在差异。采用金刚石砂轮可以进行TC4钛合金的高效深磨,相比金刚石砂轮,CBN砂轮显得更适于钛合金磨削。G. Werner建立的磨削力数学模型对TC4钛合金高效深磨也有普遍的意义。单位宽度磨削功率与单位宽度的磨粒耕犁面积存在一定的线性增长关系。基于遗传算法的进化神经网络在收敛速度和预测精度方面明显优于传统BP神经网络,且有效的避免了局部极小,在工程中有较高的实用价值。更多还原

【Abstract】 Titanium alloy are becoming popular materials in many areas for their highperformances such as high specific strength, good corrosion resistance, antimagneticcapability, high thermal endurance and fatigue resistance. But their highperformances also makes them difficult to grind. It will make some problems such ashigh abrasive force and grinding temperature, adherence, bad wearance, if somegeneral ways has been used to grind titanium alloy. So, new advanced grindingmethods are needed urgently to improve machining productivity and grindingsurface integrality.High efficiency deep grinding (HEDG) refers to an advanced modern grindingprocess, which utilizes the high or ultra-high grinding speeds to achieve theunimaginable high grinding efficiency, high grinding ratio and also satisfactorysurface quality. It has brought the traditional grinding technology to a revolutionarystage and revealed its deep hidden developing possibilities and prosperousapplication vista. It has great help about HEDG in application to the grinding oftitanium alloy for widening its range of application and searching new efficiencyways to grinding titanium alloy. Based on the experiments about high efficiencydeep grinding 40Cr steel, systemic experiments was done about grinding TC4titanium alloy under the high efficiency deep grinding condition. In the experimentsthe grinding force and workpiece surface roughness of two materials in the conditionof HEDG were measured by change the grinding conditions. The author discussedthe effect of various parameters on grinding force, specific grinding energy, grindingforce ratio, studied the indicated material removal modes and force formationmechanisms, also develop the mathematical model of the high efficiency deepgrinding force of the TC4 titanium alloy in the condition of adherence. Then,theresearch about the characteristics, formation of specific grinding energy and theconsumption of grinding power was carried out. At last, the prediction model ofworkpiece surface roughness was made by using traditional BP neural network andevolutionary neural networks. The results of using different ways was analyzed andcompared.The results indicated that the grinding conditions have great effect to unit areagrinding force and unit area grinding force, specific grinding energy corresponding to the maximum undeformed chip thickness and the equivlent cutting thickness well.The removal modes of two materials are mainly plastic and the materials areremoved by gliding and ploughing when the great grinding depth and high workpiecevelocity are applied. There has difference in gringing force and specific grindingenergy because of their mechanical properties. Diamond wheel can be used to highefficiency deep grind TC4 titanium alloy, but CBN wheel is much better in grindingit. The grinding force model of built by G. Werner is also prevalent in the TC4titanium alloy high efficiency deep grinding. The results also indicated that grindingpower per unit width monotonously increases with the grain plough area per unitwidth. Evolutionary neural networks based genetic algorithm is obviously better thantraditional BP neural network in the field of convergence rate and accuracy ofpredictions. It can effectively overcome the problem of falling into local minimumpoint and have great use value in the engineering.更多还原