Prediction Method Of Compressive Elastic Modulus Of Solid Wood Sheet Based On Fiber Angle Detection

Wood is a renewable biomass material that is widely used in construction,bridges,furniture,etc.,but its application range is limited due to lack of sufficient understanding of its mechanical properties.Compressive elastic modulus is an important mechanical index for wood applications The traditional detection method is measured by mechanical breakthrough experiments,which are characterized by harsh conditions,long time-consuming,and partial bias.With the development of modern sensing technology,the non-destructive testing method of wood mechanical strength has been vigorously developed.At present,the main non-destructive testing technologies include:near-infrared spectroscopy,density prediction,and stress wave.As an important feature of the surface of the sheet,the fiber angle determines the mechanical properties of the sheet to a large extent.By constructing the model of sheet fiber angular distribution and mechanical properties of sheet metal,it can make up for the defects of expensive measurement,complicated operation and limited methods in the current non-destructive testing of mechanical strength of mainstream wood,and expand the non-destructive testing method of sheet metal mechanics.Starting from the tracheid effect of coniferous material,a set of fiber angle detection platform integrating light source emission,spot collection and sheet traversal is designed to realize the statistical function of fiber angle measurement and fiber angle distribution.Aiming at the problem that the camera has a trapezoidal deformation of the spot image relative to the surface of the plate to be tested during the device construction process,a special mark is set on the surface to be tested to complete the posture recovery of the camera,thereby completing image correction.In the process of designing the fiber angle of the plate from the laser spot image,the Gaussian filter kernel is used to suppress the high-frequency signal of the image,and the edge of the spot is extracted by the gradient operator.In the process of extracting the maximum deformation direction of the spot,the elliptic equation is used as the model,and the least squares method is used for fitting.For the defect that the least squares is sensitive to the "outlier point",the sampling principle is adopted by using the consistency principle,and the influence of the outliers on the ellipse fitting is reduced by means of heuristic search.Due to the jitter of the pulsed laser source’s transmitting power,the measuring device has fluctuations in different subsamplings of the same point.The fluctuation conforms to the Gaussian distribution,so the mean filtering is used to smooth the final fitted fiber angle.Through the traversal of the plate to obtain the fiber angular distribution of the plate,after experimental optimization,the fiber angle mean,fiber angle standard deviation and sneak factor of the positive and negative sides of the plate were determined as input,and the compressive elastic modulus was used as the output to construct the BP neural network.The task of predicting the compressive elastic modulus of the sheet fiber angular distribution is completed.In the fiber angle measurement process,the maximum error of the device is 0.65°when the final fiber angle mean filter parameter is 20 by single-point repeated measurement.According to the national standard "GB-T-15777-1995".100 larch samples were processed,and the fiber angle distribution was collected by the testing platform to obtain the compressive elastic modulus by mechanical testing machine;the training samples and test were divided according to the ratio of 3:1.Sample;repeatedly determined by experiment,the feedforward neural network with 6-15-4-1 is constructed with the mean value of fiber angle,fiber angle variance and sneak coefficient as input,and the compressive elastic modulus is output;squared by network weight As a cost function regular term,the NAGD method is used for training,and the network accuracy can reach 90.8%,which can better complete the stress-resistant elastic modulus prediction task of the sheet.