A novel methodology has been developed for extracting pyrolysis kinetic parameters from thermogravimetric analysis (TGA) data. The development of this methodology is motivated by a need to automate the determination of material properties for use in fire models. The algorithm with which the methodology is implemented is described. Aside from being fully-automated, the resultant script has the advantage of being efficient—a full set of kinetic parameters is provided in less than one second. The script is verified against manufactured TGA data for one and two reaction mechanisms and the effects of reaction peak width and the distance between reaction peaks is examined. Validation is accomplished by applying the script to TGA data for Nylon 6,6, a flexible polyurethane (PU) foam, and polyvinyl chloride (PVC). The resultant kinetic parameters are tabulated, and plots of the actual and predicted TGA data show that the algorithm is quite effective for one, two, and three reaction mechanisms. INTRODUCTION Computational fire models have proven to be effective at predicting the spread of heat and smoke in a wide range of building fire scenarios. However, such models still generally require user input describing the actual fire that is generating the heat and smoke. Computational predictions of flame spread and fire growth require somewhat detailed models of condensed phase physics, and a number of condensed phase pyrolysis models have been developed. Such models have proven effective at modeling the burning rate of small slabs, but their application to flame spread calculations is more limited. Part of the problem is that these pyrolysis models require the specification of a large number of material properties. Furthermore, there are many different flammable materials that need to be considered in fire scenarios. Substantial progress in applying computational fire models to predicting flame spread could be achieved by the development of both a streamlined methodology for characterizing the thermophysical properties of flammable materials as well as the creation of a publicly available database of such material properties. One successful approach for characterizing materials is based on performing a number of milligramscale and bench scale tests such as thermogravimetric analysis (TGA), differential scanning calorimetry (DSC), and the controlled atmosphere pyrolysis apparatus (CAPA). An example of a public database of material properties is found in the Validation Guide of the Fire Dynamics Simulator (FDS). Currently, an international effort is underway to further develop experimental and modeling tools such as these in order to “advance predictive fire modelling”. The present manuscript presents work based on a coordinated effort to develop a comprehensive database of material properties for use in fire models. This database will include both raw data from a suite of milligramscale tests and a list of properties for each material that could be directly used as inputs for a fire model such as FDS. In order to generate these tables of material properties, automated computational scripts are required to robustly analyze raw data from small scale tests. In this paper, we present such a script for calibrating a generalized pyrolysis kinetic model to TGA data. Previous work has looked at optimization algorithms and Markov Chain Monte Carlo (MCMC) methods for fitting TGA data, but these approaches are not fully automated for general multi-step reaction schemes and can be relatively computationally expensive.

中文翻译：

---

热重分析数据的自动拟合

已开发出一种从热重分析 (TGA) 数据中提取热解动力学参数的新方法。开发这种方法的动机是需要自动确定用于火灾模型的材料属性。描述了实现该方法的算法。除了完全自动化之外，生成的脚本还具有高效的优点——在不到一秒的时间内提供全套动力学参数。该脚本针对一种和两种反应机制的制造 TGA 数据进行了验证，并检查了反应峰宽和反应峰之间距离的影响。通过将脚本应用于尼龙 6,6、柔性聚氨酯 (PU) 泡沫和聚氯乙烯 (PVC) 的 TGA 数据来完成验证。将所得动力学参数制成表格，实际和预测的 TGA 数据图表明该算法对于一种、两种和三种反应机制非常有效。引言 计算火灾模型已被证明可有效预测各种建筑火灾场景中的热量和烟雾的传播。但是，此类模型通常仍需要用户输入来描述产生热量和烟雾的实际火灾。火焰蔓延和火势增长的计算预测需要一些详细的凝聚相物理模型，并且已经开发了许多凝聚相热解模型。此类模型已被证明可有效模拟小板坯的燃烧速率，但它们在火焰蔓延计算中的应用更为有限。部分问题是这些热解模型需要对大量材料属性进行规范。此外，在火灾场景中需要考虑许多不同的易燃材料。通过开发用于表征易燃材料的热物理特性的简化方法以及创建此类材料特性的公开可用数据库，可以在应用计算火灾模型预测火焰蔓延方面取得重大进展。表征材料的一种成功方法是基于执行大量毫克级和实验室规模的测试，例如热重分析 (TGA)、差示扫描量热法 (DSC) 和可控气氛热解装置 (CAPA)。材料属性公共数据库的示例可在 Fire Dynamics Simulator (FDS) 的验证指南中找到。目前，正在进行一项国际努力，以进一步开发诸如此类的实验和建模工具，以“推进预测性火灾建模”。本手稿介绍了基于协调努力开发用于火灾模型的材料特性综合数据库的工作。该数据库将包括来自一组毫克级测试的原始数据和可直接用作 FDS 等火灾模型输入的每种材料的属性列表。为了生成这些材料属性表，​​需要自动计算脚本来稳健地分析来自小规模测试的原始数据。在本文中，我们提供了这样一个脚本，用于将广义热解动力学模型校准为 TGA 数据。以前的工作着眼于拟合 TGA 数据的优化算法和马尔可夫链蒙特卡罗 (MCMC) 方法，但这些方法对于一般的多步反应方案并不是完全自动化的，并且计算成本相对较高。