Abstract:

Objective: In order to provide a theoretical basis for further improving the accuracy of wood long-term life prediction, the applicability of the commonly used Gerhards model in wood life prediction research was analyzed, the influences of the selection of model parameters on the life prediction results were discussed, and an interval model that can better reflect the distribution characteristics of test data based on the Gerhards model also proposed. Method: Firstly, the Gerhards model was theoretically deduced to clarify the relationships between the failure time of the wood specimen and the ramp loading speed or constant stress level. Then, by comparing the real values of coniferous wood loading test data in the literature with the predicted values of 8 groups of Gerhards models, the pros and cons of each model for wood life predictions of different loading conditions were analyzed. Finally, aiming at the problem that the predicted value is higher than the true value in wood life prediction, an interval prediction model analysis method based on Gerhards model was proposed. Result: When the Gerhards model is used to predict the life of wood specimens under linear loading conditions, the predicted values are relatively close, and the deviation from the experimental values is small; when predicting the life of wood specimens under constant loading conditions, the difference between each predicted values is large, and some of the predicted value are very different from the experimental values. When the constant stress level on the Northern Douglas fir sample is between 60% and 95%, it is recommended to use the Gerhards model 6 for life prediction. When the stress level on the Norway spruce sample is between 55% and 105%, it is recommended to use the Gerhards model 7 for life prediction. When the stress level on the wood sample is lower than 50%, it is recommended to use the Gerhards model 1 for life prediction. The difference between the loading condition of the data used in the fitting model and the loading condition of the predicted object has a great influence on the model prediction accuracy. Compared with the traditional prediction models, the interval model can reflect the distribution characteristics of the test data more completely, and the interval prediction model fitted by percentiles can cover more test data than that fitted by the data based on T distribution hypothesis. Conclusion: It is necessary to discuss the applicability of the Gerhards model in the study of wood life prediction. The model could have a better prediction effect when the loading condition of the data used in the fitting model is similar to the loading condition of the predicted object. The quantile model proposed in this paper can make most of the test data points fall within the interval, which can provide a reference for improving the long-term life prediction accuracy of wood.

Key words: wood, Gerhards model, long-term life prediction, applicability analysis, interval model