Chinese plantation poplar reaches strength class C30 on stiffness, but scatter is high: full-scale grading of 1,043 boards

Introduction

Fast-growing plantation poplar is one of the most abundant hardwood resources in China, but it is mostly used for panels, packaging and non-structural products. Whether it can be graded and sold as structural lumber depends on knowing how its stiffness is distributed across a large population of boards, and on being able to assign those boards to a recognised strength class. A team from Nanjing Tech University (College of Civil Engineering) and Central South University of Forestry and Technology (Changsha), China, addresses this in a paper published in the journal Forests in September 2024 under the title "Strength Grading of Full-Scale Chinese Fast-Growing Poplar Wood for Structural Building Applications".

According to the authors, the study sets out to do four things: test whether the Weibull, normal and log-normal distributions are suitable models for the modulus of elasticity (MOE) of the poplar; assign a strength grade; examine how density and moisture content influence MOE; and judge whether the wood is suitable as a load-bearing material. The work uses full-size boards, 2.6 m long, and not small clear specimens, which makes it closer to what a sawmill actually produces than most laboratory wood-property studies.

The paper is relevant to sawmills and graders because it shows, with a large sample, what happens when a visual pre-selection and a machine-style stiffness measurement are applied to a single plantation poplar clone. All numbers below are the authors' own; the portal has not repeated the analysis, except where it says so.

Research methodology

Material. The study used 2,072 specimens of poplar (Populus euramericana cv. 'San Martina I-72/58') taken from 32 trees grown in Suqian City, Jiangsu Province. The trees were at least 12 years old. The mean annual ring width was 10.86 mm and the latewood share 35.37 percent. Boards had a cross-section of 40 x 150 mm and a length of 2,600 mm, and they were industrially dried to a moisture content of 8 to 15 percent.

Visual grading. Specimens were first visually graded according to the Chinese standard GB/T 26899. Those with defects outside the permitted limits were excluded: 1,043 boards passed and went on to mechanical testing, which means that roughly half of the 2,072 boards (1,029) were rejected before any stiffness measurement.

Measurements. Density was determined according to GB/T 1927.5. MOE was determined by three-point bending on an MTS testing machine according to GB/T 1927.10, with a loading rate of 2 mm/min, a load range of 10 to 20 kN and a span of 2,400 mm. Three load cycles were applied and the average of the last two was used. Moisture content was read with a pen-type moisture meter, three readings per board, averaged. The MOE values were then adjusted to a reference moisture content of 12 percent with a linear correction factor of 1 + 0.015(W - 12), where W is the measured moisture content.

Statistics. Normality was checked with the skewness and kurtosis Z-scores (threshold 1.96) in SPSS 26; log-normality was checked the same way on the natural logarithm of MOE. The Weibull parameters were estimated by least-squares regression and the fit was tested with a Kolmogorov-Smirnov test at the 5 percent level. Spearman correlation and a correlation matrix were used for density, moisture content and MOE.

Grading approach. The strength grade was derived from the mean MOE only, using a characteristic-value approach following EN 384, in which the mean characteristic MOE parallel to the grain is taken as the mean divided by 0.95. The authors note that the European strength-class standard EN 338 normally considers further properties, so the grade in this paper rests on stiffness alone.

Key findings

Distribution of MOE. For the normal distribution the skewness Z-score was 0.259 and the kurtosis Z-score 0.235, both below 1.96, so the MOE data were accepted as normally distributed. For the log-normal distribution the Z-scores were 8.87 and 3.38, both above the threshold, and the log-normal model was rejected. A Weibull fit gave R-squared of 0.969, a shape parameter (Weibull modulus) m of 5.1403 and a scale parameter of 12.936 GPa; the Kolmogorov-Smirnov test returned D = 0.026 and P = 0.051, just above 0.05, so the Weibull hypothesis was accepted. The margin is thin: the P-value is only marginally above the acceptance threshold.

Mean stiffness and grade. From the Weibull fit the authors report a mean MOE of 11.604 GPa and a standard deviation of 4.63 GPa. Dividing the mean by 0.95 gives a mean characteristic bending MOE parallel to the grain of 12.21 GPa. The average characteristic density was 521 kg/m3. On this basis the poplar was assigned to strength class C30. The authors attribute the higher grade, compared with earlier work, to the removal of visually defective lumber at the start.

Reliability. The Weibull modulus of 5.14 is interpreted by the authors as indicating lower reliability and uniformity. For that reason they judge the poplar unsuitable for direct structural use, even though the stiffness-based grade is C30. They describe the MOE distribution as asymmetric about 12.44 GPa. A reader should therefore keep two statements apart: the sample reaches C30 on mean stiffness, and the authors still conclude that scatter is too large for direct structural use.

Density and moisture content. The Spearman analysis showed that both density and moisture content significantly affect MOE (p < 0.01). Linear regression described the density-MOE relationship better than a power law, with R-squared of 0.511, so density alone explains only about half of the variance. In the correlation matrix, MOE correlated with density at r = 0.62 and with moisture content at r = 0.033; the authors call the density correlation relatively low. The fitted moisture-content curve did not agree well with the correction formula used; the authors attribute this to specimen size and non-uniform moisture distribution in large sections and suggest that local MOE measurement would be more accurate for large cross-sections.

Recommendations of the authors. Pre-selecting by appearance improves consistency, and selection should additionally consider density requirements, although no density threshold is given. They also advise considering several factors, not only density and moisture content, when predicting mechanical properties.

Points the portal could not reconcile. When the portal checked the reported Weibull parameters, the standard two-parameter formula gave a mean of about 11.9 GPa, close to the reported 11.604 GPa, but a standard deviation of about 2.75 GPa, which is well below the reported 4.63 GPa. The portal cannot tell whether this reflects a different fitting convention, a typographical error or an inconsistency in the paper, and the reported figure is given here as published. The Weibull density coefficient and exponent in some equations also looked inconsistent with the fitted parameters in the text version, which was partly garbled.

Practical applications for sawmills

Portal interpretation (these points are not claims made by the paper):

• Visual pre-sorting is a large filter. Roughly half of the boards in this sample were removed by the visual rules before testing. A mill planning to supply poplar for structural or semi-structural products should expect a substantial share of its output to fall out at the first sorting step and should plan the commercial use of that share.
• Stiffness measurement adds information that density cannot give. Density explained about half of the MOE variance and the correlation was moderate (r = 0.62). If a mill wants to sell stiffness-graded poplar, the paper supports measuring stiffness directly (for example by machine grading) in place of relying on density sorting alone. The paper does not test a machine-grading device; it uses laboratory bending tests.
• Reporting the mean is not enough for customers. The study shows a case where the mean stiffness gives a high class while the spread is judged unacceptable by the authors. Buyers of structural lumber typically care about the lower tail of the distribution, so a producer would need to report variability and not only the mean. The paper itself does not compute fifth-percentile values.
• Moisture at grading time matters. The correction to 12 percent worked poorly in this study. Mills drying poplar to a wide target range (here 8 to 15 percent) and grading afterwards should be aware that uneven moisture inside thick boards may blur the stiffness readings.
• Resource-use context. The paper supports the view that Chinese plantation poplar has more potential than its current non-structural use suggests, provided it is selected carefully; it does not provide an economic analysis or yield figures.

Limitations

The captured text contains no separate limitations section. The following are stated or evident from the paper:

• Only one poplar clone from one location (Suqian City, Jiangsu) and 32 trees were sampled, so the results cannot be taken as representative of poplar in general.
• Only MOE was measured. Bending strength was not tested, so the link between stiffness and strength relies on earlier literature, and the grade is not a full EN 338 classification.
• Because about half the boards were excluded by visual grading, the results describe a pre-selected population and not the raw sawn output.
• The Kolmogorov-Smirnov result for the Weibull model (P = 0.051) is only just above the acceptance threshold.
• The moisture correction fitted poorly, and the density correlation is described as relatively low.
• The reported standard deviation does not match the value that the portal computed from the reported Weibull parameters (see above).
• Several parts could not be read in the version used for this summary: Figures 1 to 5 are images (the Weibull fit plot, distribution curves, correlation diagrams and moisture-content curve), Table S1 with the 1,043 individual records is supplementary material, and some equations were garbled in the extraction. Table 1 had a garbled header row, although its values were readable. The regression coefficients for density against MOE are not stated in the text.

Conclusions

On a sample of 1,043 visually pre-selected, full-size poplar boards from Jiangsu, the stiffness distribution is consistent with normal and Weibull models but not with a log-normal model. The mean characteristic MOE of 12.21 GPa and mean density of 521 kg/m3 lead to strength class C30 on a stiffness-only basis. At the same time, the low Weibull modulus of 5.14 leads the authors to judge the material unsuitable for direct structural use. Density and moisture content both affect MOE, but density alone is a weak predictor. For the sawmill sector the value of the paper is in the combination of a large full-scale sample and a transparent statement of the pre-selection, and in the reminder that mean stiffness, variability and sorting yield must be considered together.

Original research

• Authors: Jin Qian, Kong Yue, Xiang Li, Yan Pu, Kaiyi Chen, Peng Wu, Zhongfeng Zhang
• Title: Strength Grading of Full-Scale Chinese Fast-Growing Poplar Wood for Structural Building Applications
• Journal: Forests, volume 15, issue 9, article 1602
• Year: 2024 (published 11 September 2024)
• DOI: 10.3390/f15091602
• Affiliations: College of Civil Engineering, Nanjing Tech University, Nanjing, China; College of Materials Science and Engineering, Central South University of Forestry and Technology, Changsha, China
• Licence: open access; the article page metadata lists CC BY 3.0 and the PDF states a Creative Commons Attribution (CC BY) licence

This article is an independent summary prepared by the portal; the original paper should be consulted for full methods, tables and figures.