王照秋,刘腾,李铁,等.船用柴油机热力过程实时预测模型[J].内燃机工程,2025,46(4):129-139.
船用柴油机热力过程实时预测模型
Real-Time Predictive Models of Marine Engines Thermodynamic Process
DOI：10.13949/j.cnki.nrjgc.2025.04.014
关键词:船用发动机  数字孪生模型  实时预测  变步长计算
Key Words:marine engine  digital twin model  real-time prediction  variablestep size calculation
基金项目:国家自然科学基金面上项目（52271325）
作者单位E-mail
王照秋* 上海交通大学 动力装置及自动化研究所，上海200240
中船动力研究院有限公司，上海200120 		W.Z.Q.q@sjtu.edu.cn 
刘腾 中船动力研究院有限公司，上海200120 lt@cspi.net.cn 
李铁* 上海交通大学 动力装置及自动化研究所，上海200240 litie@sjtu.edu.cn 
黄帅 上海交通大学 动力装置及自动化研究所，上海200240 zzdett@163.com 
摘要点击次数: 559
全文下载次数: 357
摘要:针对船用发动机数字孪生模型的缸内狄塞尔循环过程高实时性计算需求，提出了一种变步长显式计算的单区实时快速预测模型。模型离散化计算点，根据精度及稳定度需求分配各阶段步长，显式化模型物理方程降低计算误差，以提升计算速度。模型简化包括：压缩膨胀模型基于多变过程，改进比热比计算方式；减少燃烧分区，简化燃烧为单步化学反应，优化燃烧计算点。模型验证基于某型320 mm缸径船用中速发动机开展。结果表明，标定的峰值压力误差小于2%，预测的峰值压力误差小于3%，CA10（放热量达到总放热量10%对应的曲轴转角）与CA50（放热量达到总放热量50%对应的曲轴转角）误差均小于1.5°。模型计算耗时均在0.25 ms内，模型能够在实现准确预测的同时，满足实时性的要求。
Abstract:To meet the high real-time computation requirements for the Diesel cycle process within the cylinder of a marine engine digital twin model， a single-zone real-time rapid prediction model with variable step-size explicit calculation was proposed. To improve computational efficiency， calculation points were discretized， step sizes for each phase were allocated according to precision and stability requirements， and physical equations were explicitly formulated to minimize computational errors. The model simplification involved adopting a polytropic process and improving the calculation method of specific heat ratio for the compression and expansion model， reducing combustion zones， simplifying combustion into a single-step chemical reaction， and optimizing combustion calculation points. The validation of the model was based on a 320 mm medium-speed marine engine. The results show that the calibrated peak pressure error is less than 2%， while the predicted peak pressure error is less than 3%. Both the errors for CA10 （the crank angle corresponding to 10% of total heat release） and CA50 （the crank angle corresponding to 50% of total heat release） are less than 1.5°. The computation time for the model is within 0.25 ms， enabling the model to achieve accurate predictions while simultaneously meeting the requirements for real-time performance.
查看全文  HTML   查看/发表评论