Abstract In previous works from the literature, the prediction precision of friction models applied for pneumatic valve modeling is compared using simulations. This paper extends these works, by comparing the performance of the friction models using a real pneumatic control valve with a smart positioner. For a proper comparison, an estimation algorithm is developed and evaluated in noiseless and Monte Carlo tests. Even though the estimator is not perfect, due to the typical SNR of industrial processes and to the complexity of the models, it was considered acceptable for comparison purposes. The results indicate that the first principle models tend to be more precise than the data-driven models.

中文翻译：

---

气动阀摩擦模型参数估计及性能比较

摘要 在文献的先前工作中，使用模拟比较了应用于气动阀建模的摩擦模型的预测精度。本文通过比较使用真实气动控制阀和智能定位器的摩擦模型的性能来扩展这些工作。为了进行适当的比较，我们开发了一种估计算法，并在无噪声和蒙特卡罗测试中进行了评估。尽管估算器并不完美，但由于工业过程的典型 SNR 和模型的复杂性，出于比较目的，它被认为是可以接受的。结果表明，第一原理模型往往比数据驱动模型更精确。