Temperature based death time estimation – Optimizing methods for practical application

基于温度的死亡时间估计(TTDE)在早期PM阶段通常由法医专家完成。在许多应用案例中，存在特殊的非标准冷却条件，即马歇尔、霍尔和亨日方法(MHH)，其中不能应用MHH模型。像MHH这样的现象学方法使用带有拟合参数的直肠温度时间曲线的特别公式。相比之下，基于物理的方法，如有限元方法，求解复杂几何形状和真实初始条件和边界条件的热方程。因此，后一种类型的TTDE技术似乎最有希望应用于实际案例工作。该项目将在早先的基于有限元-TTDE结果的DFG项目的基础上，重点加强基于有限元的TTDE方法在实际案例中的适用性。从上一个项目的个性化有限元-TTDE开始，我们希望在较少模型参数的情况下研究简化的车身冷却数学模型。由于它们可以用更少的时间、IT和测量设备进行参数化，因此可以在更大的范围内直接应用于犯罪现场案件工作TTDE。新项目将研究以下三种为TTDE生成简化模型的方法：-代理模型-动态简化模型-几何简化模型用于代理模型的方法包括解析表示法和神经网络。利用有限元生成的训练数据，通过机器学习完成模型的构建。简化模型反过来又是具有特定于问题的基函数的离散化。几何简化模型通过省略或粗化远离核心温度测量点的模型区域的有限元网格来缩小个性化的有限元模型。上述所有简化模型生成类型都是从前一个项目产生的个性化有限元模型开始的。在代理模型的情况下，这是通过使用个性化有限元模型来生成所需的训练数据来完成的。进一步的目标是：-表示身体和环境之间复杂的热力学相互作用。-对环境情景及其相关热力学过程及其要记录的热力学参数进行分类--(部分)重建人体冷却期间的环境温度时间曲线，并以简化方法加以实施。气候室可以用来测量热力学数据，并模拟上面提到的针对幻影或真实人体的研究的真实冷却情景。热相互作用将使用红外相机捕捉，物体表面将通过3D相机进行数字化。

Temperature based death time estimation (TTDE) in the early pm phase is usually done by forensic specialists. In many application cases there are special non-standard cooling conditions in the sense of the Marshall and Hoare and Henßge approach (MHH), where the MHH-model cannot be applied. Phenomenological methods like MHH use ad hoc formulae of the rectal temperature time curve with fitted parameters. In contrast, physics based approaches like the Finite-Element-Method (FEM) solve the heat equation for complex geometries and real initial- and boundary-conditions. Therefore, the latter type of TTDE-technique seems most promising for real case work application. The project shall focus on enhancing the applicability of FEM-based TTDE-methods in practical casework on the basis of the earlier DFG-project on FEM-TTDE’s outcome. Starting from the individualized FEM-TTDE of the last project, we would like to investigate simplified mathematical models of body cooling with a smaller number of model parameters. Since they can be parametrized with much smaller effort in time, IT-, and measurement-equipment, direct application in crime scene case work TTDE will be possible on a much larger scale. The new project shall investigate the following three approaches to generate simplified models for TTDE: - Surrogate models - Dynamically reduced models - Geometrically reduced models The approaches used for Surrogate models include analytical representations as well as neural networks. Model construction is done by machine learning with FEM-generated training data. Reduced models in turn are discretizations with problem-specific basis functions. Geometrically reduced models downsize the individualized FE-model by omitting or coarsening the FE-mesh of model regions far from the core temperature measurement point. All of the aforementioned simplified model generation types start from individualized FEM-models resulting from the predecessor project. In case of the surrogate models this is done by using the individualized FE-model to generate the training-data necessary. Further goals are: - Representation of complex thermodynamic interactions between corpse and environment. - Classification of ambient scenarios and of their relevant thermodynamic processes as well as of their thermodynamic parameters to be recorded - (Partial) reconstruction of ambient temperature time curves during body cooling and their implementation in simplified methods. A climatic chamber can be used to measure thermodynamic data and mimic real cooling scenarios for the investigations mentioned above for phantoms or real human bodies. Thermal interactions will be captured using an IR camera, and object surfaces are digitized by 3D camera.