Many experiments are usually needed to quantify probabilistic fatigue behavior in metals. Previous attempts used separate artificial neural network (ANN) to calculate different probabilistic ranges which can be computationally demanding for building probabilistic fatigue constant life diagram (CLD). Alternatively, we propose using probabilistic neural network (PNNs) which can capture data distribution parameters. The resulted model is generative and can quantify aleatoric uncertainty using a single network. Two tests are presented. The first captures the fatigue life aleatoric uncertainty for P355NL1 steel and successfully builds a probabilistic fatigue CLD. The resulted network is not only more efficient but also provides higher accuracy compared with ANN. To assess fatigue, the second test examines vibrations of a pipework assembly. The proposed methodology quantifies the nonlinear relation between the vibration velocity and the equivalent stress and successfully reflects measurements uncertainties in fatigue assessment. The proposed methodology is published in opensource format (https://github.com/MShadiNashed/probabilistic-machine-learning-for-fatigue-data).

The risk of vibration-induced fatigue in process pipework is usually assessed through vibration measurements. For small-bore pipework, integrity personnel would measure the vibration of the pipework and refer to widely used charts to quantify the risk of vibration-induced fatigue. If the vibration levels are classified as OK, no action is required on behalf of the operators. However, if it is a CONCERN or PROBLEM vibration level, strain measurements are required to adequately quantify the risk through a fatigue life assessment. In this paper, we examine the suitability of a widely used vibration acceptance criteria through finite element models. A total of 4,800 models are used to study the suitability of this vibration acceptance criteria by monitoring both the vibration and dynamic stress. The model comprises a small-bore pipe (2″ SCH 40) that is fitted on a mainline size 5″ SCH 40 using a weldolet; the length of the mainline takes three values resulting in three models. The mainline supporting conditions will be varied using translational and rotational springs. The finite element models will be excited using a point load resembling flow-induced forces (with varying flow velocity and fluid composition). These excitations are obtained from the literature and are based on experimental studies as power spectral density functions. The results show that the studied vibration acceptance criterion is suitable in 99.73% of all the studied models with 68.27% confidence level. For the models with a shorter mainline pipe, the criterial is suitable in 76.5% of the time with 68.27% confidence level.

No abstract