【作者】 刘登峰；

【导师】 徐保国；

【作者基本信息】 江南大学 ， 控制理论与控制工程， 2014， 博士

【摘要】 黄酒是我国特有的传统酒种，其发酵过程目前仍主要靠人工控制来实现，造成批次质量的稳定性较差。随着产能及质量需求的提升，迫切需要具有高鲁棒性的控制系统来实现其过程自动化，而构建准确描述发酵过程特性的数学模型是建立自动控制系统的基础，黄酒发酵过程是典型半固半液发酵，其发酵系统极其复杂，且缺乏足够的发酵动力学数据，目前鲜有研究涉及黄酒发酵过程的建模。本文以黄酒酿造过程为研究对象，首先通过生物实验确定关键阶段，并进一步采集了关键阶段的详细过程数据，然后以实验数据为基础分别对糖化过程、发酵过程以及同时糖化和发酵的双边过程进行了建模与参数辨识，具体研究内容如下:(1)针对决定黄酒品质的关键阶段不确定尤其是缺少过程数据的问题，本文通过实验模拟黄酒发酵过程，基于高效液相色谱技术(HPLC)测定了不同酶和曲在不同温度梯度下前、后酵结束两个关键点酒醪中主体成分含量，利用SAS分析了产物浓度的差异性，结果表明黄酒主体成分主要在前酵阶段生成。另外通过实验获得了不同温度下黄酒前酵过程数据，确定了主要状态变量，为建模及模型简化提供了理论及数据支持。(2)针对曲中酶的种类及比例不确定，提出模型结构来解决该问题。利用高效薄层层析技术(HPTLC)确定糖化过程中主要寡糖分布，并利用HPLC技术研究糖化过程中寡糖的浓度变化，基于所得的数据以低阶为目的建立了适合单酶、双酶和黄酒前酵糖化过程的模型结构。并基于四阶龙格库塔法和加权最小二乘算法利用模拟实验测得的数据对该模型进行了验证，结果表明所建模型能准确跟踪寡糖的变化。(3)针对黄酒醪液中酵母难以分析的问题，以葡萄糖为底物，系统研究了不同温度和不同起始底物浓度下酵母Su-25的动力学变化。并分别利用典型的Hinshelwood模型和Monod模型对该过程进行了仿真验证，结果表明Monod模型更适合描述Su-25发酵的动力学过程。(4)基于前期建立的糖化模型和发酵模型，构建了以寡糖、乙醇和溶氧等为状态变量的黄酒前酵双边发酵过程(SSF)模型，并分别利用所采集的实验室水平和工厂实际生产过程数据进行了模型辨识与验证。结果表明模型在实验室水平和工厂实际生产水平均能很好描述黄酒酿制中同时糖化和发酵的双边过程，模型能实现对实验数据的预测。(5)针对传统反馈控制或在线控制技术难以直接应用于包括黄酒在内的间歇发酵过程，本文提出基于广义预测控制的思想，通过优化黄酒发酵初始条件来实现优化的策略，该优化可以看作是系统控制时域为1而预测时域为整批反应时间的预测控制。通过设定预期目标轨迹，加入实际过程中变量浓度限制条件，利用最小二乘法来进行酿造条件的优化。仿真结果表明该策略可以实现黄酒发酵酿造条件的优化。本文所获得的实验数据对黄酒工业生产具有实际的指导意义和参考价值，所构建的模型对实现黄酒发酵过程自动化、提高批次稳定性具有重要的理论意义和应用价值。更多还原

【Abstract】 Rice wine is one of the oldest alcoholic beverages in the world. The primary phase ofrice wine fermentation is a typical simultaneous saccharification and fermentation (SSF)process and is also referred to as a semi-solid state and semi-liquid state fermentation process.This process is critical to rice wine quality control, but there has been little published work onkinetic modeling and control of glutinous rice saccharification and rice wine fermentation. Inthis research, glutinous rice saccharification and fermentation were studied experimentallyand theoretically to develop process models and control strategies. To gain insights into theinfluential system variables, a sequence of experimental studies were carried out to determinethe influence of fermentation temperature and source of enzymes on the ecologicalcharacteristics of rice wine, and the effect of temperature on Chinese rice wine brewing withhigh-concentration pre-steamed whole sticky rice. After kinetic models were developedseparately for glutinous rice saccharification and fermentation, a SSF process model wasdeveloped for the rice wine production process.Rice wine samples were produced with four sources of enzymes at three fermentationtemperatures. Enological variables, including ethanol, main sugars, glycerol, and organicacids, were measured by HPLC at the end of primary fermentation (4days) and at the end ofpost fermentation (40days). The data showed that both source of enzymes and temperaturehad significant effects on the concentrations of the measured variables. The results provideinsights into the rice wine fermentation process as affected by different enzymes andfermentation temperatures.The effects of fermentation temperatures on Chinese rice wine quality were investigated.The compositions and concentrations of ethanol, sugars, glycerol, and organic acids in themash of Chinese rice wine samples were determined by HPLC. The experimental resultsindicated that temperature contributed significantly to ethanol production, acid flavor contents,and sugar contents in the fermentation broth of the Chinese rice wines.Glutinous rice saccharification was performed by using α-amylase, glucoamylase, two-enzyme combination, or wheat qu. Experiments were carried out at two different locations,with rice from different sources, and in varied fermentation temperatures. The main productswere identified and measured by HPLC. Low-order kinetic model structures (forms orconstructs of model with adjustable parameters) were proposed based on the major chemicalreactions brought about by different enzymes. The model structures were then tested for theirabilities to capture the main kinetic variations after parameter optimization by a least-squaresalgorithm. The proposed model structures were found useful in representing measured kineticvariations except those in maltotriose produced with wheat qu. The estimated reactions ratescorrectly reflected the variations observed from the experiments and provided insights into thereaction processes in terms of reaction speeds, dominant variations, and primary products. Theactions of α-amylase and wheat qu differed from findings in prior research. The proposedmodel structures show promise for describing the saccharification process of glutinous rice.A kinetic model structure was developed for the fermentation process by Su-25based on the biochemical reactions involved. Experiments with the Chinese rice wine yeast underdifferent conditions were performed and used to validate the model structure. It was foundthat the model structure could decribe the ferementation process. The developed modelstructure can be used to control or optimize rice wine production.Rice wine fermentation was performed by using pre-steamed rice, Chinese wheat qu andrice wine yeast strains Saccharomyces cerevisiae Su-25. Experiments were carried out withdifferent conditions, and the main products were identified and measured by HPLC. A low-order kinetic model structure was proposed based on the major chemical reactions in the ricewine fermentation process. The model structure was then tested for its abilities to capture themain kinetic variations after parameter optimization by a least-squares algorithm. Theproposed model structure showed promise for describing the fermentation process of ricewine.Another kinetic model structure was developed for the fermentation process of Su-25according to the involved biochemical reactions. The model structure can be used for differenttemperatures and different initial substrate concentrations. Experiments with Chinese ricewine yeast under different conditions were performed and used to validate the model structure.The model structure was verified with experiments in both the lab scale and the plant scale.Wine fermentation is a batch processes, for which conventional feedback controltechniques are either ineffective or inapplicable. Based on the framework of GeneralizedPredictive Control (GPC), a predictive control strategy was formulated to control the batchfermentation process by varying the initial process conditions. The formulation is equivalentto GPC with a unit control horizon but a long predictive horizon and its implemented boilsdown to least-squares optimization of an initial process condition vector for a given processoutput target under variaous constraints. Simulations were performed to demonstrate that thetechnique could drive the process towards desired output targets.更多还原