Establishing acoustic-based measurement protocol

Transcription

1 Deliverable D.4.05 Establishing acoustic-based measurement protocol WP4 Multi-sensor model-based quality of mountain forest production Task 4.4 Data mining and model integration of log/biomass quality indicators from stress-wave (SW) measurements, for the determination of the SW quality index Revision: Final Authors: Mariapaola Riggio Author name CNR Dissemination level Contributor(s) Reviewer(s) Editor(s) Partner in charge(s) PU (Public) Stefano Marrazza (COMPOLAB) Federico Prandi (GRAPHITECH), Jakub Sandak (CNR) Raffaele De Amicis (GRAPHITECH) CNR Due date Submission Date final submission Page 1 of 45

2 REVISION HISTORY AND STATEMENT OF ORIGINALITY Statement of originality This deliverable contains original unpublished work except where clearly indicated otherwise. Acknowledgement of previously published material and of the work of others has been made through appropriate citation, quotation or both. Page 2 of 45

3 Table of contents REVISION HISTORY AND STATEMENT OF ORIGINALITY... 2 Statement of originality... 2 Table of contents... 3 List of figures... 4 Acronyms Introduction Application of stress wave-based techniques in forestry and wood characterization Material-dependent factors affecting acoustic measurements Anisotropy and heterogeneity Moisture content Temperature Methodology-dependent factors affecting acoustic measurements Frequency Transmission modes Coupling Determination of log quality indicators from stress-wave (SW) Defectiveness indicators Slope of grain Compression wood Knottiness Decay indicators Insect decay Rot Density/stiffness indicator Stress-wave data acquisition and analysis Test setup Signal processing and data mining Protocol of acoustic measurement within SLOPE for the determination of the quality indicators Preliminary analysis: SW measurement on standing trees Preliminary analysis: SW measurement on trees after felling SW analysis on de-branched logs Measurement of visible defects Direct measurement of wood material properties correlated with SW data 29 Page 3 of 45

4 4.5.1 Factors affecting correlation between direct measures and SW data 32 5 Determination of SW quality index Stress-wave velocity conversion models Stress wave and relation with measurement position in the stem Stress wave and relation with log diameter Incorporation of parameters from different types of measurements Stress wave (SW) quality indexes Stress wave (SW) quality index # Stress wave (SW) quality index # Test plan for on-line SW measurement in the processor head Laser measurement system Accelerometers measurement system References List of figures Figure 1 : Types of stress-waves... 6 Figure 2: Orthotropic axis in a tree trunk: global (L,R,T) and local (1,2,3)... 9 Figure 3: Different transmission modes. a) direct longitudinal, b) transverse, c) indirect longitudinal, d) semi-direct Figure 4: Different head-mounted sensor tips tested Figure 5: Sensors array for the measurement of the slope of grain on standing trees (Bucur, 2006). E (emission point); ϕ (grain angle); F (fiber longitudinal axis); G (geometric axis of the tree); α (measurement angle); R (receiving points) Figure 6: Spruce bark beetle attack Figure 7: Severe rot decay in a tree trunk ( 20 Figure 8: Portable test setup for SW data acquisition and analysis (TDAS 16 - Boviar s.r.l.) Figure 9: laboratory test setup for SW data acquisition Figure 10. Definition of the measuring points on the standing trees Figure 11. Schematic of the clear wood samples extracted from the log discs Page 4 of 45

5 Figure 12. Diagrammatic radial variation of density in spruce (breast height) (from Smith et al., 2003) Figure 13. Basic stem-growth diagrams showing the distribution of the density in Spruce (from Kollmann et al., 1984) Figure 14: Concept of the quality map indicating different SW velocity along the log Figure 15: Concept for the laser measurement system integrated in the processor head Figure 16: Concept for the placing and decoupling system for accelerometer Figure 17: Concept of the accelerometers positioning system on the processor head Figure 18: On the left the accelerometer positioning system in rest position; on the right during stress wave measurement Acronyms EMC FSP MC MoE MoR NDT P-waves S-waves SW ToF US Equilibrium Moisture Content Fiber Saturation Point Moisture content Modulus of Elasticity Modulus of Rupture Non-destructive testing Pressure or compression waves Shear waves Stress wave Time-of-flight ultrasound Page 5 of 45

6 1 Introduction Stress waves are generated into the wood when a stress is applied suddenly to its surface. Three types of waves are initiated by this impact, two types of bulk waves and surface waves (Rayleigh waves) (Figure 1). Longitudinal waves (or compression or P-waves) correspond to particle motions parallel to the wave propagation, while transversal waves (or shear or S-waves) correspond to particle motion in a direction perpendicular to the wave propagation. While most energy resulting from an impact is carried by surface and S-waves, P- waves are the fastest and the easiest to detect in field application. Figure 1 : Types of stress-waves Main descriptors of acoustic waves are velocity, rate of energy flow and attenuation. These change during propagation, as influenced by the characteristics of the medium. Acoustic velocity can be calculated from the span between two sensing points (S) and the so called time-of-flight (ToF) t: VV = SS tt [1] In the resonance method, a stress wave is initiated by an impact on one end of the log and the acoustic velocity measured is a weighted average of the acoustic pulses resonating longitudinally in the log. In this case the acoustic velocity can be determined from f 0, the fundamental natural frequency of the acoustic wave signal, and the log length L. VV = 2ff 0 LL [2] Page 6 of 45

7 1.1 Application of stress wave-based techniques in forestry and wood characterization Techniques based on the measurement of stress wave ToF are the most explored acoustic non-destructive methods for evaluating the quality of standing trees in forests [Ross et al. 2000, Yin at al. 2011]. In applications on stranding trees, measurements are done in the lower part of the stem (generally at the breast height), using impact stress waves. Measurements done along the length of the trunk are generally aimed at evaluating the mechanical quality of the material (pre-grading). In order to extend the results in a specific location (breast height) to the whole stem, prediction models have been developed by many researchers (e.g. Haines and Leban, 1997; Wang et al., 2000). These models are based on the consideration that the systematic variation in wood properties such as module-of-elasticity (MoE), module-of-rupture (MoR), density, are both species specific and influenced by silvicultural practices (Auty, D. and Achim, A., 2008, Achim 2009). One commercial device developed at this scope is the HITMAN ST300 (http: Correlation of stress wave propagation parameters on standing trees and on corresponding butt logs is not straightforward. Wang et al deduced that, in standing trees, propagation is dominated by dilatational or quasi-dilatational waves rather than one-dimensional plane waves. In the cited research, variability in velocity prediction is reduced using both dilatational wave models and multivariate regression models. Transversal measurements (across the trunk) are generally aimed at detecting inner defects and decay. The sonic tomography (SoT) technique, based on the determination of ToF of impact stress waves, is one of the available techniques, to map internal decay in standing trees [e.g. Socco et al. 2004]. Commercial devices such as PiCUS ( ARBOTOM (http :// d=11 ) and FAKOPP3D ( are specifically dedicated to application of acoustic tomography in the forestry sector. The resonance method is generally considered more reliable than ToF data (. It cannot be applied on trees, but it can be used for the characterization of logs and in general, it is an advantageous technique to be applied in case of long slender members, with a cylinder-like shape and know boundary conditions (e.g. free-free beam). Page 7 of 45

8 1.2 Material-dependent factors affecting acoustic measurements Anisotropy and heterogeneity Wood can be considered as an orthotropic elastic solid, or alternatively as cylindrically anisotropic medium. When bulk waves propagate in an isotropic solid, particle motion along the propagation direction form longitudinal waves, while motion perpendicularly to the propagation form shear (transverse) waves. In case of anisotropic materials both shear and longitudinal waves can propagate either along the principal directions or out of them (Bucur, 2006). This phenomenon must be taken into account for the stress-wave investigation of wood, in order to avoid misinterpretation of travel time. Wood anisotropy at the different macroscopic levels affect the acoustical behaviour of wood. At the cellular level, wood structures can be considered as a system of closed tubes embedded in a matrix. The tube/cell walls provide a continuous wavepath. In the longitudinal direction, the dissipation of acoustical energy takes place at the limit of the tubes. In the radial direction, acoustic waves find another type of conducting tubular structure (the medullary rays), while this is not present in the tangential direction. In the radial direction, the layered structure of the annual ring, with thin-walled earlywood cells and thick-walled latewood cells, behaves like a filter with alternating pass bands and stop bands. This is particularly evident in species such as spruce, with rings composed of layers with very different density values (earlywood density is in the average less the half of latewood density). The continuum theory, ignoring the layered structure of wood can be adopted only if the wavelength is long compared with the annual ring thickness (Bucur, 2006) Table 1 presents the values of ultrasonic waves (1MHz) in the symmetry axes of small clear samples of softwoods conditioned at 12% moisture-content (MC), as reported by Bucur, Page 8 of 45

9 Table 1. Average values of ultrasonic velocities (m/s) of longitudinal waves (V 11, V 22 and V 33 ) in the three anatomical directions (L, R, T) and shear waves (V 44, V 55, V 66 ) Species Density Ultrasonic velocities (kg/m 3 ) V 11 V 22 V 33 V 44 V 55 V 66 Silver Spruce* 352 5,500 2,225 1, ,386 1,361 Spruce* 400 5,600 2,000 1, ,425 1,374 Douglas fir 440 5,500 2,330 1, ,660 1,622 Stika 430 5,550 2,300 1, ,480 1,500 spruce* Common 450 5,200 2,200 1, ,560 1,630 Sitka spruce Red spruce 485 6,000 2,150 1, ,240 1,320 Common 485 5,353 1,580 1, ,230 1,322 spruce Pine 580 5,000 2,100 1, ,030 1,050 *species used for musical instruments (from Bucur, 2006) At the level of the tree trunk, the anisotropic organization of the material can be represented by two systems of axis, as in Fig. 2, (from Brémaud et al. 2011): the [R, T, L] system aligned to the pith of the tree and a local system [1, 2, 3] fitting the grain s orientation. In case of negligible stem taper and no source of fibre deviation, both systems of axis can be assumed coincident. Figure 2: Orthotropic axis in a tree trunk: global (L,R,T) and local (1,2,3) Page 9 of 45

10 In softwoods, however, spiral grain develops after the initial growth of the tree, reaching a maximum value in the first ten years. After this period, often the grain angle decreases towards zero, and then spiral direction is inverted. In Picea abies the initial spiral is left-handed and inversion of spirality occurs in an extremely variable time (Harris, 1988). Brémaud et al studied the effect of grain angle on the anisotropy of wood vibrational properties. They found that, in softwoods, the axial-to-transverse ratio of dynamic elastic moduli range from 2.2 to 4.4 while the axial-to-shear anisotropy for damping coefficients ranges from 1.4 to 3.3. A solid can be considered homogeneous for wave propagation if the wavelength of the signal in at least one direction is bigger than a tenth the encountered inhomogeneity. Inhomogeneous elements in wood are present at different scales, from the cellular scale to the macroscopic scale. The presence of macroscopic inhomogeneity in wood can limit the accuracy of stress-wave readings. For example, if the size of the internal discontinuities is comparable with the pulse wavelength, the pulse is attenuated by scattering at interfaces. Changes in the measurements of stress-wave parameters in different zones can indicate the degree of inhomogeneity of the tested material Moisture content The fiber saturation point (FSP), that is the moisture content at which only the cell wall are completely saturated with water, is considered the status above which wood properties do not change as a function of MC. In trees and freshly cut logs, moisture exists in wood, both as free water in cell lumina and as bound water within cell walls. Therefore MC is well above FSP. With a MC below the FPS, the cellulosic material in the cell wall or the cell boundaries are the main causes of attenuation. With MC above FSP, wood porosity is a predominant factor in scattering. If stress-wave methods are used to grade timber, it is important to take into consideration that MC affects the two input parameters in the calculation of the dynamic MoE: density and sound velocity. Some researchers proposed adjustment of these properties to the reference condition at 12% equilibrium moisture content (EMC). Sandoz 1993 and Steiger 1995 reported moisture correction factors for the ultrasonic ToF measurements in wooden boards below and above the FSP. Unterwieser and Schickhofer (2011) analysed influence of moisture content on Page 10 of 45

11 the estimation of dynamic MoE from both natural frequency and ultrasonic ToF data. The Authors concluded that while moisture adjustments below FSP can be accomplished with linear functions, which are applicable for both methods. In the case of MC above FSP the two methods require a different approach: while velocity of ultrasound from ToF exhibited a linear dependence on MC, stress-wave velocity determined from the natural frequency (resonance method) exhibits no MC dependency. Stress-wave velocity data collected during the validation phase of the task 4.4 in the SLOPE project will be evaluated considering the differing influence of MC on the measurements. From an operational point of view, moisture content in wood influences the applicability of a specific stress-wave methodology. In fact, moisture content above FSP causes high attenuation of the signal, thus hindering application of ultrasonic methodologies, since high frequency signals are more susceptible to attenuation and signal loss. For this reason, the application of stress-wave measurements in SLOPE will be focused only on the use of low frequency signals, initiated by the impact of a hammer-like device Temperature For the acoustical evaluation of trees and harvested logs in mountainous areas, it is important to consider the different climates at the time of testing. In fact, wood temperature can range from above 30 C to below the freezing point (0 C), depending on geographical locations and the harvesting season. Gao et al observed that ambient temperature has a significant influence on stress wave velocity in winter, when the temperature is below the water freezing point. This is attributed to the phase change of free water in the cell cavities, and the consequent higher acoustic velocity in the frozen material (velocity in water is 1482 km/s at 20 C and in ice is 3800 km/s). In the cited paper, a linear regression model is proposed, for adjusting the reading results at the different temperature conditions, as expressed in Eq. [3]: 11 CC TT 2.5 CC VV = aa + bbbb 2.5 CC TT 0 CC 0 CC TT 30 CC [3] where the regression coefficients a and b are given for red pine trees (Gao et al. 2013). Authors also recommend to avoid measurement at temperature around freezing, being temperature adjustment less reliable in this condition. Page 11 of 45

12 1.3 Methodology-dependent factors affecting acoustic measurements Frequency In acoustic measurements, the frequency of the transmitted signal is one of the parameters that have to be carefully selected, in order to obtain a reasonable compromise between the signal attenuation and resolution. High frequencies are more sensitive to internal defects, but attenuation will worsen with increasing frequencies. If attenuation is too severe, the signals can be completely lost or used only for measurements on short spans. In the application foreseen in the SLOPE project, the main sources of attenuation are the high moisture content of the green wood, the presence of the bark, the length of the wave path (distance between the transducers). Other possible sources of attenuation are the presence of defects and decay in the inspected logs. For these reasons, low frequency signals will be used to energize the logs to be tested. In the foreseen integration of the SW apparatus in the processor-head, an impact hammer will produce the low frequency signals. The material of the hammer head changes the frequency of the produced wave; harder materials produce higher frequencies stress waves, while softer materials induce lower frequency. Also the weight of the hammer can affect the frequency; heavy hammers induce lower frequencies than light hammers since they experience a longer time of contact with the impacted surface. In the SLOPE project, the characteristics of the hammer (material of the hammer head, weight) and of the design of the mechanical release system will be defined in the further planning phase. It has been observed that that the dynamic MoE, estimated from stress-wave methods depends on the frequency used. In particular, dynamic MoE increases with frequency (so that the lowest MoE value is the static one). This property is shared by all solid materials and finds its theoretical explanation in the Kramers Kronig dispersion relations (Ouis, 2002). The causes of dispersion can be explained by considering wood as a highly damping and viscoelastic material. As such, the dissipative force is proportional to the velocity and the restored elastic force is proportional to the displacement. This phenomenon will be taken into account during the validation phase of the Task 4.5 of the SLOPE project, when properties will be measured directly (e.g. by static loading tests) to develop the models for the definition of the SW quality index. Page 12 of 45

13 1.3.2 Transmission modes The definition of the sensing layout depends on the accessibility of the longitudinal or transversal faces of the logs and the desired path of stress wave travel. Depending on the position of the sensors, stress wave propagation will be measured in either a) direct longitudinal, b) transverse, c) indirect longitudinal or d) semi-direct mode (Figure 3). According to the mode of stress wave transmission, the sensor (and eventually the screws or spikes at the sensor head) should be connected either perpendicular to the surface, for direct longitudinal, transverse and semi-direct wave propagation, or at an angle of approximately 45, and maximum 60, with the timber surface in case of indirect longitudinal wave transmission. It must be taken into account that, when using embedded transducers, the angle between the transducer and member can affect the transit time if it is too large and should be considered when testing. Figure 3: Different transmission modes. a) direct longitudinal, b) transverse, c) indirect longitudinal, d) semi-direct If the stress wave is generated by indirect impact on a longitudinal face of the log (mode c of Figure 3), large part of the energy is absorbed and dissipated in a three-dimensional content. In case of small log diameters the wave may travel with the log as a quasi-plane wave, for large diameters, however, the propagation phenomenon is dominated by dilatational waves (Wang et al. 2007). In the latter case the stress wave velocity is dependent on both modulus of elasticity and Poisson s ratio. The higher velocity values, generally observed from ToF measurements in indirect transmission mode (longitudinal readings on the trunk) have been also attributed to the higher stiffness of the outerwood portions (Chauhan and Walker 2006). In many species this is dependent on the tree age Coupling One of the main limitations of traditional stress-wave sensors is often connected with the contact and the coupling required to have a satisfactory signal-to-noise ratio. In application on wood, the problem of coupling is connected to the roughness of the tested surface (for instance, if the bark is present). To overcome this problem, spikes or screws are often connected to the sensors used in the non-destructive evaluation of standing trees. In this case, the spikes penetrate into the bark and transmit the signal to the sapwood layer. In some applications, especially for in situ NDT evaluation of timber structures, coupling media are used, such as couplants gel, which are viscous, have a good acoustic impedance Page 13 of 45

14 and can be used for rough surfaces. The use of a given coupling medium also depends on the environmental conditions at the moment of testing, basically on temperature and humidity. Glycerine, mineral greases, silicone resins and wax are generally used at room temperature. Alcohols are sometimes used at low temperatures. It must be noted, however, that the use of coupling medium can produce unexpected errors in the measures (due to the penetration into the wood, the excessive thickness of the couplants layer). In the SLOPE application, the use of coupling media is not very realistic. Given the necessity to automatize (and mechanize) the measurement process. As an alternative to contact methods, some non-contact techniques have been developed based on air-coupled ultrasound or, in some cases on immersion techniques. In the frame of the SLOPE project the problem of the coupling from the sensors to the tested logs have been deeply considered. The specific testing conditions, which affect coupling, are: the presence of the bark the limitations of the mechanized system, namely: o fast insertion of the pins into the trunk/log, under the bark, and, consequently fast and safe pull-out of the pins); o necessity of de-coupling from the machine, or alternatively to filter out machine noise, to avoid superimposition of vibrations. Two options have been therefore considered and tested: 1) non contact measurements with a laser displacement sensors; 2) contact measurements with contact accelerometers, with different types of head-mounted tips/spikes. For the first option, a LR-T reflective sensor (Keyence, com/products/sensor/photoelectric/lr-t/index.jsp) has been tested. Good preliminary results have been obtained in the laboratory setup. However, also some limitations of the single linear laser beam have been highlighted, such as the possibility to encounter to light traps, in measurements on the longitudinal faces, such as cavities and slots into the bark, which hinter the readings. A possible solution to this problem is the use of a laser profilometer, which however is generally a costly device. If measurements are done on the, freshly cut transversal face, the single linear laser can be used, without encountering with shadows problems. The second option foresees the use of traditional accelerometers (20 khz frequency). In this case, the quality of the readings, using different type of Page 14 of 45

15 pins, has been evaluated in the lab (Figure 4). It has been observed that a better coupling is achieved for sharper pins and screws. Further tests will be necessary to define the optimal geometry of the pins, which can be also apt to be used during the automatic measurement with the processor. In the case of contact measurement, the problem of the interference with the machine vibrations (head processor) have to be faced. Possible technical solutions will be studied. Figure 4: Different head-mounted sensor tips tested Page 15 of 45

16 2 Determination of log quality indicators from stress-wave (SW) 2.1 Defectiveness indicators Stress waves propagation is affected by the anatomical structure of wood, therefore specific anatomical features of the material, as well abnormalities of these features can be potentially detected by acoustic methods. The structural elements of the wood material and their organization can be described at different scales (i.e. from the nano scale of the cellulose molecule, to the micro scale of the wooden cells, up to the meso level, including growth rings). Size of structural wood elements is much smaller than the wavelength λ used in acoustic-ultrasound experiments (3-10 cm). Theoretically, the frequency field used in the investigation must be related to a wavelength that is comparable with the dimensions of the investigated wood structural element, which vibrates as elementary resonator Slope of grain Bucur and Perrin 1990 proposed to estimate slope of grain, both on standing trees and on samples, on the base of three measurements of longitudinal velocities as in Figure 5. In the SLOPE project, the possibility to detect of slope of grain in the tested material will be explored. For this purpose, different testing configurations (i.e. reciprocal position of the probes) will be tested, both in the indirect longitudinal transmission mode and in the semi-direct mode. Page 16 of 45

17 Figure 5: Sensors array for the measurement of the slope of grain on standing trees (Bucur, 2006). E (emission point); ϕ (grain angle); F (fiber longitudinal axis); G (geometric axis of the tree); α (measurement angle); R (receiving points) Compression wood Compression wood is a gravitropic reaction of softwood trees (gymnosperms). Specific environmental and climatic conditions, such as the exposure of the site to strong wind and heavy wet snow loads can cause the formation of compression wood, while the slope has not been proven to be an influencing factor [Duncker and Spiecker, 2005]. In compression wood, tracheids are shorter than in normal wood and fibril angle is higher. Higher specific gravity is associated with elevated lignin content. The velocity of acoustic waves, in all anisotropic directions, is lower in reaction wood than in normal wood [Saadat-nia et al. 2011]. In the cited researches, stress wave measurements were done on stem disks, analysing velocity variation in the different radial positions. Acoustic tomography has been proposed for the nondestructive detection of compression wood in standing trees [Brancheriau et al. 2012]. Page 17 of 45

18 In the SLOPE framework, measurements will be carried out on standing spruce trees characterized by a significant change of stem curvature below the breast height, in order to explore the possibility to detect compression wood. Measurements are planned in the pilot stand at Piscine Sover, Sover municipality (TN), using impact stress waves. Slices will be then cut from the tested portion of the tree and analysed in the laboratory Knottiness Gerhards (1982) studied the influence of knots in stress waves propagation. He suggested that the presence of knots and related grain deviation cause the wave front to diverge from the normal direction, as theorized for long slender rods, thus resulting in a lower wave speed. This problem is particularly evident in large, short wood members. In this case, radial inertia and wave interaction with the external, free surfaces of the material play a relevant role (Wang et al. 2004). The SW testing procedure developed in the SLOPE project will include both estimation of the overall quality of the log (and therefore of the level of defectiveness distributed in the whole log) and evaluation of the local quality of portion of the log, for optimization of the logging process. The former analysis will be based on measurements on the transversal faces, adopting the resonance method (longitudinal vibration). The latter analysis will be based on comparison of global and local velocity values, these measured with the ToF method. 2.2 Decay indicators Insect decay An insect, typically attacking spruce trees is the so-called European spruce bark beetle (Ips typographus L). It normally attacks already suffering plants. The insects create pitch tubes under the bark (Figure 6). The depreciation of the wood is an indirect effect of the attack, caused by the infestation of beetle-associated bluestain fungi, such as the Ophiostoma and Ceratocystis species. Also the great spruce bark beetle (Dendroctonus micans) damages spruce trees by tunnelling into the bark. In all these types of attack, the boundary between the bark and the trunk contains air, larvae, and other low-density matter. Acoustical investigations of infested logs (with bark) can highlight the damage, given the high attenuation of the signal caused by the discontinuities between the bark and the wood. Page 18 of 45

19 2.2.2 Rot Figure 6: Spruce bark beetle attack Decay can be present in trees at different stages. Stem rot infects tree through pathways formed by wounds on the stem or on broken branches. If the conditions are appropriate, colonization spreads to the surrounding wood. Physical and chemical barriers produced by living trees are defensive mechanisms adopted by the nature to prevent the spread of decay. One of the early warnings of fungal attack, before any visible evidence of discoloration and decay, is the increase of moisture content in wood. Fungal activity affects moisture content in wood by the production of metabolic water, the transport of free water from the sapwood, and, indirectly, by the capillarity effect caused by the different pore volume. In standing trees, however, moisture content is well above the fiber saturation point, and estimation of higher decaydependent moisture levels is not trivial. In case of extended decay, the material will loose is consistency, and, in extreme cases cavities will form. Cavities can be sensed acoustically even by simple tapping on a stem with an axe or a hammer. Indeed, acoustic waves are attenuated more in decayed wood, with damages rotten parts or cavities (Figure 7). Page 19 of 45

20 Figure 7: Severe rot decay in a tree trunk ( The application of stress-wave techniques for the detection of decay in trunks or logs is based on the analysis of the wave propagation in the radial-tangential plane. Some manufacturers of acoustic tools provide some threshold values of the relative velocity decrease in percentage, to estimate the degree of decay (as. e.g. decayed area ratio). The relative decrease in stress-wave velocity can be expressed as below: ΔVV = VV rrrrrr VV VV rrrrrr 100% [4] An example of such a guideline is given in Table 2, as reported by the Fakopp Microsecond timer user s guide, Table 2. Decay estimation from compared decrease of SW velocities Relative Velocity decrease in % Decayed area ratio in % 0-10 No decay >50 >50 However, the detection of fungal decay can be influenced by the presence of other sources of heterogeneity. Page 20 of 45

21 Several researches observe that a single path stress wave is only able to identify decay, if it covers more than 20% of the cross-sectional stem area (Nicolotti et al. 2003, Ross et al. 2004). 2.3 Density/stiffness indicator The velocity of propagation of a bulk p-wave in an infinite isotropic solid is related to the elastic constants and the density (Equation 5): VV LL = EE ρρ = λλ+2μμ ρρ = [5] where E is the dynamic modulus of elasticity, ρ is the density, λ and μ are the socalled Lamé coefficient, respectively the shear modulus and the wave length. On the basis of this relationship it is possible to estimate the mechanical properties (stiffness) of the tested material. The mechanical characterization of the material is not within the scopes of the SLOPE project. However, the screening of the velocity values of the tested logs will be used for pre-grading logs. Therefore, high quality logs (low defectiveness, higher stiffness) will be characterized by high velocity values, and conversely, lower grades will be characterized by low velocity values. Page 21 of 45

22 3 Stress-wave data acquisition and analysis 3.1 Test setup In the SLOPE project, stress-wave data will be acquired at the different stages of the harvesting chain: on the standing trees, on felled trees and on the debranched logs. The test setups will differ in the three scenarios: I. in the application on both standing and felled trees, a portable equipment will be used, constituted by (Figure 8): a. a multi-channel device, allowing signal generation and preliminary analysis. It is connected to a PC for data storage and signal processing; b. an instrumented hammer for emitting signals in the sonic frequencies; c. an ultrasonic probes, 55 khz frequency. A screw is mounted to the probe head, in order to pierce bark and cambium and extend into sapwood, and reduce the contact area between the probe and the specimen; d. consistent coupling is achieved by manual pressure. Figure 8: Portable test setup for SW data acquisition and analysis (TDAS 16 - Boviar s.r.l.) Page 22 of 45

23 II. For testing de-branched logs, the test setup will be mounted on an ad hoc scanner, under construction (Figure 9). The test set up is constituted by two modules: 1) To analyse the longitudinal axial modes, with the resonance method: a. an impact hammer or a pendulus, impacting on the transversal face of the log; b. a laser displacement sensor, with the laser point directed toward the transversal face of the log; c. a data acquisition system (National Instrument Model NI CRIO and NI9233 four channels). 2) To analyse the velocity of p-waves in transverse, indirect longitudinal, and semi-direct transmission modes: a. an impact hammer or a pendulus, impacting on the longitudinal face of the log; b. two accelerometers, positioned on the longitudinal/trasversal face of the log, with the reciprocal position varying according to the specific transmission mode (see Fig. 3), the various acceleration time histories are collected with a sampling ratio of 20 khz.; c. a data acquisition system (National Instrument Model NI CRIO and NI9233 four channels). The laboratory scanner is designed by the Consortium, in particular by the CNR team, in collaboration with COMPOLAB. The scanner will be installed at the IVALSA premises, and will be used for extensive testing all the functionalities foreseen for the acquisition of sensor data from the processor. Figure 9: laboratory test setup for SW data acquisition Page 23 of 45

24 3.2 Signal processing and data mining In acoustic methods, a burst or pulse is imparted into the material and the response is captured at another point. From signal processing of the received signal, different types of signal waveform parameters can be extracted. We know that the received signal is a result of wave interactions, multiple reflections and mode changes (Beall, 2002). Different parameters, measuring velocity, frequency content or attenuation, can describe different characteristics of the signal. A list of possible parameters, for the stress-wave characterization of wood, is reported by Senalik et al In particular the following parameters have been highlighted in the literature, as providing the minimum of redundancy and maximum of information (Beall, 2002): the root mean square (RMS) voltage, used to measure the signal energy; the signal velocity (V); the time centroid (TC), which is related to the frequency shape In the SLOPE project, the characterization of the wood material will be based on the measurement of the signal velocity (V). In dispersive media, such as wood, the stress-wave velocity is dependent on frequency, and both group and phase velocity can be measured: group velocity (velocity of signal energy/wave packet) phase velocity (computed from the phase delay/time of flight) The relationships between phase velocity V and group velocity v can be summarized by the equation: vv = VV 1 ff dddd [6] VVdddd In the SLOPE project the phase velocity will be calculate from the time-of-flight data and the resonance method. Time-of-flight can be measured in different ways. Most commercial ToF devices are based on the simple voltage threshold detection method. The threshold level must be set just above the noise level. The method, however, is very sensitive in variations of signal amplitude. In dispersive materials, the leading edge of the waveform may be difficult to detect. A more robust method is the so-called slope-detection. In this case, ToF is measured from the time difference between the two start points of the first slopes in the signal (above noise threshold). Page 24 of 45

25 In the SW protocol developed in T4.4, a cross-correlation function of the signals from the two transducers will be used, to improve the accuracy of the ToF estimates, from the data acquired with the scanner setup (setup II-2). As suggested by Emms and Hartington, 2014, a calibration of the receiving transducers will be performed, by measuring both the time delay (d 1 ) from a source closest to the first accelerometer and (d 2 ) from a source closest to the second accelerometer. Therefore, the calibrated ToF measurement t, will results in: tt = dd1 dd2 2 [7] A time window is defined, to allow both long signal duration and undistorted signal shape. According to Emms and Hartington, 2014, a time window out the first complete pulse is effective, to accurately pick up the p-wave signal, avoiding reflected signals or surface waves. Also, a Fourier interpolation will be used to increase the signal time base resolution. The velocity of the stress waves will be determined from the time delay t and the distance between the two transducers. Statistical regression analysis will be undertaken following either a stepwise multiple regression methodology and a simple linear least squares regression analysis. This will allow examination of the correlation between the phase velocity values determined with the different testing procedures, and the properties (physical, mechanical, anatomical) directly measured. The significance of the explanatory variables will be determined, and threshold values will be indicated for these variables, according to the specific scopes of the pre-grading procedure established. Page 25 of 45

26 4 Protocol of acoustic measurement within SLOPE for the determination of the quality indicators 4.1 Preliminary analysis: SW measurement on standing trees A series of tests are on-going at the pilot stand in Sover (TN, Italy), to evaluate the variables affecting the SW readings on standing trees, and as a consequence, on the material tested in the SLOPE project. The procedure followed in this testing phase is the following: a) Identification of the stand b) Identification of the tree (local conditions) c) Measurement of air humidity, air temperature d) Definition of the measuring points Test height Test span Orientation of the measuring points around the circumference of the stem (according to the scheme in Figure 10) a) Removal of the bark at the location of the measuring points b) Identification of gross defects in the inspected area (defined as the linear path between sensors and transmitter impact point-) c) Measurement of the DBH (diameter at breast height 130 cm) d) Signal acquisition test setup defined in Chapter 3.1 The span between the two receiving points (h) and the position of the transmitting signal (impact) is varied in this phase, in order to evaluate the different influence of the testing layout. Page 26 of 45

27 Figure 10. Definition of the measuring points on the standing trees 4.2 Preliminary analysis: SW measurement on trees after felling This phase will be initiated with the starting of the harvesting activity in the pilot stands in Sover (TN, Italy). It aims at verifying the correspondence of the measurement on the standing tree and on a log or trunk after felling. The procedure that will be followed in this testing phase is the following: a) Identification of the tree (RFID tag) b) Definition of the date of felling c) Measurement of air humidity, air temperature d) Definition of the measuring points i. at the point already measured on the standing trees ii. random position across the diameter e) Measurement of the trunk diameter at the mid position between measuring points Page 27 of 45

28 f) Measurement of the distance between receiving points g) Measurement of the distance between receiving points and transmitting point (impact) e) Signal acquisition test setup defined in Chapter SW analysis on de-branched logs This testing campaign will be intensively carried out in the CNR labs, using a prototype scanner. This simulates the measurements that will be performed with the processor head, once the integration phase (Task 6.4) will be completed. Logs coming from the demo areas constitute the main test material. The logs will be transported immediately after felling at the CNR premises, and a first series of tests will be carried out. These measurements are done at moisture contents close to those of logs processed in the forest. Another series of tests will be performed on the same logs, at different moisture contents, while they are conditioning, in the open service area at the CNR premises. The following data will be acquired on the de-branched logs: 1) velocity of the longitudinal P-waves with the resonance method 2) velocity of the quasi-longitudinal P-waves with the ToF method (indirect and semi-direct modes) 3) velocity of the transversal P-waves with the ToF method (direct transverse mode) The general protocol for the preliminary analysis is the following: a) Identification of the tree (RFID tag) b) Indication of the date of felling c) Position of the log along the tree: it is defined as the distance from the breast height d) Measurement of air humidity, air temperature e) Measurement of the wood moisture content with the resistance hygrometer (for MC values below 30%) f) Measurement of the log diameters (end diameters, diameters at the measuring points) g) Measurement of the log length (according to EN /2) h) Measurement of the log weight i) Measurement of the visible defects in the logs (according to EN , EN 1310) Then, according to the acquisition method: j) Definition of the measuring points k) Signal acquisition test setup defined in Chapter 3.1 Page 28 of 45

29 4.4 Measurement of visible defects The analysis of macroscopic defects on the tested material is a step included in the measurement protocols, as reported in the previous paragraphs. The scope of this analysis is to verify the influence of specific defects, or macroscopic heterogeneity, on the investigated SW data. The features that will be measured are those, described in the D 4.01 Existing grading rules for log/biomass, which are downgrading factors of the quality class of Norway spruce logs. Therefore, the following features will be measured, according to the EN 1310 standard: Knots According to the EN standard, the knot types are distinguished in: Sound knot Dead or Loose Knot Unsound knot Resin pocket Spiral grain Eccentric pith Compression wood (if visible) Sweep Shakes Checks Splits Insect or worm holes Rot The extension and position of the above-mentioned features with regard to the position of the sensing points will be recorded in the testing reports. 4.5 Direct measurement of wood material properties correlated with SW data In this phase, some parameters and characteristics of the tested material, which can be correlated with the measured SW data, will be determined in the laboratory with direct methods (i.e. density from weight/volume data, MoE from bending tests, etc.). The aim of this phase is to build a database of material properties, in particular those that are relevant for log grading (i.e. strength Page 29 of 45

30 grading) and define the relationship between these properties and the SW-NDT variables. In particular, the following properties will be determined: Apparent density of the green log, ρ l : ρρ ll = WW ll VV ll (kg/m 3 ) [8] where W l and V l are the green weight and green volume of the log, respectively. The log volume and weight are calculated according to the EN standard. Apparent density of the log at varying MC, ρ l [%] : ρρ ll[%] = WW ll[%] VV ll[%] (kg/m 3 ) [9] where W l[%] and V l[%] are, respectively, the weight and volume of the log at a given MC. The log volume and weight are calculated according to the EN standard. Density of green clear wood samples, ρ w : ρρ ww = WW ww VV ww (kg/m 3 ) [10] where W w and V w are the green weight and volume of clear wood samples, respectively. The samples are prismatic elements, without macroscopic defects, extracted from discs of the tested logs, according to the schema in Figure 11. Volume and weight of the clear wood samples are calculated according to the EN 408 standard. Page 30 of 45

31 Figure 11. Schematic of the clear wood samples extracted from the log discs. Density of conditioned clear wood samples, ρ [12%] : ρρ ww[12%] = WW ww[12%] VV ww[12%] (kg/m 3 ) [11] where W w[12%] and V w[12%] are, respectively, the weight and volume of clear wood samples, conditioned at 12% moisture content. The samples are the same used for the density measurements in green conditions, as described in Figure 11. Volume and weight of the clear wood samples are calculated according to the EN 408 standard. Modulus of elasticity in compression parallel to the grain of conditioned clear wood samples, E c,0 : The modulus of elasticity in compression will be determined according to the EN 408 standard. The test pieces will be extracted from samples used for density determination (Figure 11). According to the EN 408 standard, the length of samples has to be six times the smaller cross-sectional dimension. Page 31 of 45

32 The modulus of elasticity in compression is determined from the data of the load/deformation plot of the compression test, according to the Eq. 12: EE cc,0 = ll 1(FF 2 FF 1 ) AA(ww 2 ww 1 ) (N/mm 2 ) [12] where F 2 -F 1 is an increment in load on the regression line of the load/deformation plot w 2 -w 1 is the increment of deformation corresponding to F 2 -F 1 Compression strength parallel to grain, f c,0 : The wood samples used for the determination of the modulus of elasticity will be also tested until failure, to determine the compression strength parallel to grain, according to the Eq. 13 (EN 408): ff mm = FF mmmmmm AA (N/mm 2 ) [13] Factors affecting correlation between direct measures and SW data It is worthwhile to specify that the prediction of mechanical properties of the harvested material is beyond the scope of the SLOPE project. Therefore the direct measurement of mechanical properties of the test material, as described in the previous section does not aim at the determination of prediction models, but, rather, to establish ranges of stress-wave velocity data for different quality indexes. Practical factors affecting the relation between stress wave velocity and modulus of elasticity are, among others: a) the deviation from the theoretical assumption of a isotropic, homogeneous, prismatic rod; b) the reliability of the density values, used in Equation 5 (if directly or indirectly measured); c) the influence of moisture content; d) the deviation of the dynamic MoE from the static one (depending on the frequency of the signal) (Divos and Tanaka, 2005); e) the variability of the material (note: the stress wave measurements are done on the whole log and MoE is determined on the samples); f) operational characteristics of the different methods used. Page 32 of 45

33 These factors have to be carefully evaluated, in order to correctly interpret data from direct measurements. Below some conclusion from previous researches are reported and will be taken into consideration, in the analysis of the experimental data in the SLOPE activity. Rais et al (2014) investigated the correlations of the dynamic MoE calculated at the different stages of the production chain, from standing trees to sawn timber, and the effectiveness of the calculated dynamic MoE values to strength grade the sawn timber. While the ToF method was used to determine the dynamic MoE of standing trees (using the Director ST300 TM ), the resonance method was used for logs and sawn timber. While the density was estimated and considered constant in the case of standing trees, it was directly measured in the other cases. Moisture adjustments were done, according Unterwieser and Schickhofer (2012). In the above-mentioned research, the relationships between the SW measurements on standing trees and both logs and boards were the lowest, (r 2 =0.37) and (r 2 =0.17), respectively. This low correlation is most probably due to the use of different methods for the determination of the MoE, and the assumption of constant values of density for standing trees. In the same research, the coefficient of determination between short logs and boards was Dynamic MoE of timber is affected by the moisture content. Most authors report a constant change in dynamic MoE for changes in MC below FSP (Sandoz, 1993; Unterwieser and Schickhofer 2011). In case of logs in green conditions, many authors report increase of dynamic MoE calculated from ultrasonic measurements and the density of wood in green condition, with increasing moisture content (Wang and Chuang, 2000). However this observation disagrees with the data from static tests, where the decrease of all mechanical properties of wood is observed with increasing moisture content. According to Wang and Chuang (2000) the dynamic MoE is relatively constant above the FSP, as all the mechanical parameters determined with static tests, and the apparent discrepancy with static tests is related to the mobility of free water. Page 33 of 45

34 5 Determination of SW quality index 5.1 Stress-wave velocity conversion models In the further development of the task 4.4 SW conversion models will be developed taking into account different variables affecting the analysis. In particular the following aspects will be analysed: Stress wave and relation with measurement position in the stem In general, many properties of wood vary along the stem. As an example, density is higher in the butt log, while the upper portion of the stem contains wood of lower density (Kollmann & Côté, 1984). According to the literature, however, the correlation between height in the tree and density is less pronounced in spruce. It must be noted that in spruce trees, the wood of lowest density is always produced near the pith, the highest density values are generally found in sapwood regions with narrow annual rings, while in mature wood of very old trees it can be possible to find wood with a slight decrease of density (Figure 12, adapted from Smith et al. 2003). Figure 13 shows a stem-growth diagram for Spruce with the distribution of the density of wood over the plane determined by the stem axis and the north-south direction (Kollmann & Côté, 1984). In the SLOPE project, the distribution of stress wave data over the stem height will be analyzed, together with the distribution throughout the cross-section of the stem. If necessary, stress-wave velocity data will be normalized on the base of the position of the reading with regard to the height in the log. Page 34 of 45

35 Figure 12. Diagrammatic radial variation of density in spruce (breast height) (from Smith et al., 2003) Figure 13. Basic stem-growth diagrams showing the distribution of the density in Spruce (from Kollmann et al., 1984) Stress wave and relation with log diameter Wang et al observed that, especially for some species (e.g. Douglas fir), dynamic MoE, estimated from direct longitudinal stress wave transmission testing, deviates progressively from static MoE as log diameter increases. A possible reason is that in high diameter logs the portion of mature wood is bigger Page 35 of 45

36 and stress waves will propagate faster through the high MoE zones. This possible phenomenon will be taken into account during the analysis of acoustic data of logs, in the SLOPE project. If necessary, stress-wave velocity data will be normalized on the base of specific log diameter ranges Incorporation of parameters from different types of measurements The peculiar aim of the SLOPE project is the integration of different parameters acquired from multiple sensors, to obtain more reliable quality indicators of the harvested material. Many factors influence SW propagation in wood, as stated in the previous chapters. Therefore, parameters measured with the other NDT methods implemented in the frame of the SLOPE project, such as log dimensions, estimated density, estimated presence and rate of decay and defects, etc., will be incorporated in the SW prediction models. Multivariate analysis algorithms (e.g. Multiple linear regression analysis, PLS) will be implemented for the definition of the importance of the different parameters (regression t-values) for the model. 5.2 Stress wave (SW) quality indexes The expected outputs of the stress wave data mining are two quality indexes associated to the local wood properties and to global properties of the log. These indicators will be appropriately combined/weighted within Tasks T4.06 to define the overall quality index (grade) for all logs harvested according to the SLOPE scenario Stress wave (SW) quality index #1 The SW quality index #1 will be in the range from 0 to 1 and will reflect the estimation of the wood quality as related to the acoustic velocity determined from the fundamental natural frequency of the acoustic wave signal, measured with the resonance method. The SW quality index #1 is a global parameter, which characterize the entire log under test and includes influence of both geometrical features and material properties. The list of variables to be included in the prediction model will incorporate (beside of the stress wave velocity): log geometry (i.e. diameter, taper, non-linearity of the stem, ellipsoid shape); position of the measurement, with respect to the log (from the butt end to the tip); Page 36 of 45

37 presence of any apparent defects (pith eccentricity, reaction wood, decay, knots, cracks, etc). The SW quality index #1 will be, whenever possible, associated to the estimated value of the wood density. The prototype of the statistical model will be developed firstly in the laboratory, to be later transferred to the processor control system. The model will consider also the effect of varying wood moisture content and acquisition-related issues, such as coupling/decoupling of sensors Stress wave (SW) quality index #2 The SW quality index #2 will be in the range from 0 to 1 and will reflect the estimation of the wood quality as related to the acoustic velocity determined from the time of flight measured with the accelerometers on the log longitudinal face. The SW quality index #2 gives an estimation of local quality, related to the portion of the tested log between the two accelerometers. The list of variables to be included in the prediction model will incorporate (beside of the stress wave velocity): log geometry (i.e. diameter); position of the measurement, with respect to the log (from the butt end to the tip); presence of any apparent defects (in particular: slope of grain, knots, decay and cracks). All these may be combined together (multi-sensor approach) in order to increase the reliability of the quality index estimation. A foreseen output of the SW quality index #2 is a quality map of log to be used for the final grading assessment and/or assisting operator in cutting decision. The concept of such map is presented in Figure 5, and includes marking of different velocity zones along the log length demonstrating presence of defects (low velocity zones) or higher quality material (high velocity zones). Figure 14: Concept of the quality map indicating different SW velocity along the log. Page 37 of 45

38 6 Test plan for on-line SW measurement in the processor head Among the others, one of the main goal of the SLOPE project is the integration of a novel intelligent system on existing harvesting machines, in order to measure wood characteristics, providing real time quality indicators, and guarantee traceability of the harvested material (both logs and biomass). Stress wave based technology has its special advantages in the field of wood nondestructive testing and quality evaluation (Hai Lin Feng, 2010) and for this reasons, implementation of above mentioned technologies in the processor head should be very important for real time grading of forest product. Following indication that arise from laboratory experiment, conducted on the setup shown in Figure 7, two systems for stress wave test will be studied and designed for a reliable integration on the harvesting machine: 1. laser measurement system; 2. accelerometers measurement system. 6.1 Laser measurement system Laser measurements are needed for longitudinal axial modes analysis with the resonance method. This system will be composed by a tool exerting on the transversal side of the tree, a mechanical pulse and a laser triangulation sensor able to detect and measure the induced free vibration of the tree (Figure 15). In order to reduce the noise produced by the vibration of the whole machine, the laser sensor should be mounted on a dumping support that reduce/eliminate the displacement due to vibration. Due to space constraints, in the processor head it should be possible that the system that will be designed for log tagging (with RFID tag) will be used as the source of the mechanical pulse. Results of such measurement system in the physical processor head should be not enough good and reliable due to the following different factors: 1. amplitude of vibration induced by the machine should be of the same magnitude, or greater than, the accuracy of the laser triangulation system despite the effect of dumping support designed; Page 38 of 45

39 2. a tool able to provide a significative magnitude of the mechanical excitation should have dimensions greater than the space constrain in the processor head. Figure 15: Concept for the laser measurement system integrated in the processor head 6.2 Accelerometers measurement system This system is necessary for wave pressure time of flight measurements in the processed wood. The system will be composed by two accelerometers (number of measured axis, accelerometer type and model will be defined during lab experiments) and a sensorized hammer that provides the mechanical stimulus and trigger signal. For this measurement system, decoupling between machine and sensors is important in order to cut out/reduce the effect of the vibration produced by the machine on the accelerometers. A possible solution is shown in Figure 16. Page 39 of 45

40 Figure 16: Concept for the placing and decoupling system for accelerometer The system shown in Figure 16 is composed by: 1. accelerometer, 2. upper support, 3. lower support, 4. decoupling actuator, 5. placing actuator. The lower support is nail shaped in order to ensure the contact of the accelerometer with the wood also in presence of the bark. On the processor head, two different accelerometers are assembled as shown in the following drawing (Figure 17). Page 40 of 45

41 Figure 17: Concept of the accelerometers positioning system on the processor head Figure 18: On the left the accelerometer positioning system in rest position; on the right during stress wave measurement Measurement procedure should be integrated in the debranching working cycle as follow: 1. Delimbing arm contracted: a. the two pistons move down placing the nail of the accelerometers in the wood; b. decoupling actuator mechanically disconnect the accelerometers; c. sensorized hammer hit the tree; d. data acquisition and measurement; e. decoupling actuator mechanically connect the accelerometers; Page 41 of 45