不确定条件下混装生产精准作业的平滑自适应调度研究

The production performances having been optimized transiently by deterministic scheduling are unstable during execution because the real production continually deviates from nominal schedule by unavoidable disturbances and noises. Thus within mixed-model assembly lines, the operation process precision (OPP) is worsen and the quality of final product is oscillated..For the objective OPP on mixed-model assembly lines, smooth adaptive scheduling is proposed for dealing with uncertainties by unifying each kind of uncertain factors with probability distribution, constructing piecewise linear feedback strategy based on the size of uncertainties and minimizing the production adjustment cost within each piece. The proposed methodology is realized from four aspects:. ① To interrupt the propagation of uncertainties and reduce the computational effort by decomposing the production scheduling process into two-level framework, with the upper level production planning and lower level operation scheduling;.② To absorb uncertainties from supply chain/ shop floor by robust optimizing the upper/ lower level respectively with proactive scheduling;.③ To inhibit the influence from known uncertain events by making adjustments about affected operations under piecewise linear feedback strategies with reactive scheduling; .④ To eliminate the influence from unknown uncertain events on the spot by analyzing and predicting uncertainties based on the integration of real-time closed-loop data collection and Bayesian principle..Based on the smooth adaptive scheduling framework, the evolution principle of cellular automata is deduced for holding uncertainties, and the existed intelligent automata scheduling model is extended so as to accelerate the computation and satisfy the real-time requirement from industries..If successful, the potential uncertainties inside and outside of the enterprise can be smooth adaptive handled during the real production so as to ensure the efficiency, stability and precision of each assembly operation, improve the quality of final products, and decrease the cost within manufacturing process.

在混装生产过程中，确定型调度方案固定不变，而噪声和扰动频繁发生。不确定事件使得确定型调度失效，引起车间作业紊乱和产品质量异动。.本项目面向多产品混流装配生产，以不确定条件下维持作业精准度为目标，利用随机变量归一化描述不确定事件，提出可分段、线性反馈不确定影响且生产调整代价最小的平滑自适应调度方法。关键从四个层面实现：①将精准作业调度分解为具有时间和空间尺度、含投产计划与作业调度的两层优化框架，中断不确定事件的传播；②基于主动调度，在投产计划与作业调度层吸收不确定事件扰动；③基于反应调度，在投产计划与作业调度层抑制已知类型不确定事件的影响；④基于实时采集数据，运用贝叶斯方法推断未知不确定事件类型。在此基础上分析元胞进化机理，研制平滑自适应调度智能算法，快速化求解以满足企业需求。.考虑不确定处理的平滑自适应调度，可确保动态环境下高效、稳健和精准的装配作业，促进产品质量的改善和制造成本的降低。

在混合装配生产过程中，确定型调度方案固定不变，而噪声和扰动频繁发生。不确定事件使得确定型调度失效，引起车间作业紊乱和产品质量异动。针对多类产品混装生产调度，以不确定条件下精准作业为目标，研究了不确定影响可控且生产调整代价最小的平滑自适应调度方法、模型和算法。主要包括：①将混装生产精准作业调度分解为具有时间和空间尺度、含投产计划与作业调度的两层优化框架，在基于多目标帕累托优化的投产排序和基于问题特性的装配线平衡研究基础上，构造了能同时考虑平衡和排序性能的越界距离评判指标，提出了基于耦合关系的平衡排序协同优化框架。②基于主动调度，构建了面向静态不确定情形的平衡排序鲁棒对等转换模型，实现了对不确定扰动的吸收，以较小的效率损失换取了生产稳定性的显著提升；通过对不确定事件的影响范围分析和相关参数更新，实现了混装生产自适应重调度，有效抑制了具有突发属性的、多类型不确定事件的影响。③考虑到混装生产的多约束、高耦合、不确定等属性，研制了面向大规模多工位装配线平衡问题的规则组合方法、融合启发式规则与群智能的优化算法、基于邻域特性的高效局部搜索算法等，在较短时间生成综合性能较优的投产计划及操作分配方案。. 本项目提出了混装生产平衡排序协同优化数学模型、面向有界或随机不确定处理的鲁棒对等转换模型、U型装配线平衡随机规划问题的线性转换方法等模型或方法，所提出的规则组合方法在5秒内找到了大规模案例的最优或近优解，群智能算法更新了2/22个第I/II类双边装配线平衡问题国际最新解。在国内重要刊物发表/录用论文49篇，其中三大检索源刊38篇，SCI检索源刊14篇，EI源刊24篇，SCI-2区以上6篇，SCI-1区3篇，授权发明专利2项，培养硕士7名、博士4名和青年教师5名。. 本项目成果可协助企业进行操作分配、投产计划编制和产线效率评估，实现动态环境下高效、稳健和精准的装配作业，促进产品质量的改善和制造成本的降低，提升装配企业智能化生产管理水平。