基于精益理念的铁皮枫斗颗粒生产全过程协同控制方法研究

Collaborative control of the entire production process of Dendrobium officinale granules based on lean principles

关键词:  铁皮枫斗颗粒  生产过程  图论算法  遗传算法  原料-工艺协同控制  过程性能

Key words:  Dendrobium officinale granule  pharmaceutical process  graph theory algorithm  genetic algorithm  raw material-process collaborative control  process performance

摘要:

目的 通过高效管理工业历史数据,从中挖掘生产规律并指导生产,从而提高产品质量的批间一致性。方法 针对中药原料异质性高、工艺参数调节灵活性不足等问题,以铁皮枫斗颗粒制药过程为研究对象,综合应用休哈特控制图、贝叶斯网络、图注意力网络、图卷积神经网络及遗传算法等先进分析工具,创新性地提出并在全生产过程中实施了原料-工艺协同控制策略。结果 该策略显著提升了水分、灰分、粗多糖和总皂苷四个关键指标的过程性能。通过预设原料质量属性的可行区间,四个关键指标的过程性能指数()均能达到六西格玛水平,其中水分、灰分和粗多糖的显著提升(水分:1.847 → 2.667；灰分:5.214 → 7.111；粗多糖(以葡萄糖计):1.889 → 2.192)。进一步研究表明,基于原料属性的工艺参数前馈控制策略应用于特定测试集批次时,四个关键指标的也能达到六西格玛水平,其中水分、灰分和粗多糖的同样得到显著提升(水分:2.052 → 5.687；灰分:7.296 → 32.934；粗多糖(以葡萄糖计):2.172 → 4.237)。 结论 所提方法可将关键成品质量属性稳定维持在六西格玛水平,且无需调整工艺流程,增量成本有限,展现出良好的工程适用性。

Abstract:

ABSTRACT:OBJECTIVE In the evolution of botanical drug from digital to intelligent pharmaceuticals, efficiently processing data, uncovering implicit production patterens, and guiding production remain pivotal research areas in botanical drug manufacturing. METHODS This study address challenges such as the high heterogeneity of raw materials and the limited flexibility in adjusting process parameters, focusing on the pharmaceutical process of compound Dendrodium granules. By employing advanced analytical tools, including Shewhart control charts, Bayesian networks, graph attention networks, graph convoloutional neural networks, and genetic algorithms, an innovative raw material-process collaborative control strategy was developed and implemented. RESULTS This strategy markedly enhanced the process performance of four key indicators: moisture, ash content, crude polysaccharides, and total saponins. By establishing feasible intervals for raw material quality attributes, the process performance indices (P_pk) of the four key indicators all reached the six-sigma level. Among them, the P_pk of moisture, ash content, and crude polysaccharides was significantly enhanced (moisture: 1.847 to 2.667; ash content: 5.214 to 7.111; crude polysaccharides: 1.889 to 2.192). Subsequent investigations revealed that the implementation of a feedforeward control strategy, based on raw material attributes and applied to specific test set batches, the P_pk of the four key indicators can also reach the six-sigma level, and the P_pk of moisture, ash content, and crude polysaccharides was also significantly improved (moisture: 2.052 to 5.687; ash content: 7.296 to 32.934; crude polysaccharides (expressed as glucose): 2.172 to 4.237). CONCLUSION The proposed method effectively maintains key quality attributes of the finished product at the six-sigma level without incurring additional costs or necessitating process modifications, thereby demonstrating substainal engineering applicability.