Integration of high-quality harvester data and new log scaling technology for efficient control of wood flow in German wood supply chains Dirk Jaeger 1, Martin Opferkuch 1, Siegmar Schönherr 1, Thilo Wagner² 1 Chair of Forest Operations University of Freiburg ²Director of Centre of Forestry Education (Forest work and Forest operations) Precision Forestry Symposium 5 th of March 2014 Albert-Ludwigs-Universität Freiburg What is the problem? Mechanized timber harvesting is increasing in Germany 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 2 1
share in total harvested volume [%] Number of harvesters 12.03.2014 Introduction Forest area in Germany: 11.1 Mio. hectares Annual Timber harvest volume about 55 Mio. m³ Mio. m³ Year 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 3 Introduction Development of mechanized harvesting in Germany About half of the annual cut is done by mechanized harvesting fully mechanized [%] motor manual/semi-mechanized [%] number of harvesters 120 100 80 60 40 20 0 1800 1600 1400 1200 1000 800 600 400 200 0 year Source: KWF (estimation) 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 4 2
What is the problem? Mechanized timber harvesting is increasing in Germany The wood supply chain is composed of many different entities 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 5 Introduction CTL wood supply chain with different stakeholders involved Source: Forest Energy Portal 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 6 3
Introduction Gap of production data Harvesting Processing Bucking Extraction Forwarding Piling at roadside Transport Mill intake 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 7 What is the problem? Mechanized timber harvesting is increasing in Germany The wood supply chain is composed of many different entities Data of harvested timber at roadside is lacking Quality control of extraction process Base for pay of harvester and forwarding contractors Planning of hauling process Quality control of delivery process to mill Documentation for forest owner Missing data is manually generated, very time consuming and costly 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 8 4
What is the problem? Manual measurement of pile volume Section method for logs 2.5 to 6.0 m in length; output is volume in stacked cubic metre (rm); section lengths vary from 1 to 10 m depending on length of pile 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 9 What is the problem? Manual measurement of pile volume End face method for logs 2.5 to 6.0 m in length; output is volume in cubic metre; distribution of log sizes, average log volume spacing of assessment lines 0.5 to 2.0 m measurement of end face diameters (u.b.) along assessment lines on front and rear side of pile 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 10 5
Goal of this study To evaluate a method for easy and accurate assessment of log volumes at roadside Solution strategy Using high quality of harvester data operating to newly established quality standards (output average log volume per product) combined with Counting exact number of logs per product piled at roadside using the photo-optical method PolterLuchs Deriving total volume of each product at roadside 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 11 Introduction Questions of interest: 1. Harvester 1.1 How accurate is the average log volume derived by a harvester calibrated to standard compared to mill measurements? 1.2 What are potential factors influencing accuracy of diameter and/or length measurements of the harvester 2. PolterLuchs 2.1 How accurate is the log number of log piles assessed by this method 2.2 What are potential factors influencing accuracy of derived log numbers 3. How accurate are derived harvest volumes by harvester/polterluchs at roadside compared to manual and mill assessments of volume 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 12 6
Introduction Introduction - Problem definition - Solution strategy Material and Methods - Quality standard for harvester - PolterLuchs - Trial set up Results - Average log volume of harvester compared to mill measurements - Potential factors influencing accuracy of harvester data - Quality of log count by PolterLuchs - Potential factors influencing log count accuracy - Accuracy of derived log volumes at roadside compared to mill data 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 13 Material and Methods Quality standard for harvesters Standard for calibration of measurement features of harvester head was developed by KWF Standard is mandatory for harvesting contractors in Nordrhein-Westfalen and Rheinland-Pfalz Basic calibration requirements for accurate harvester measurement standard for harvesters by KWF 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 14 7
Material and Methods Lastenheft Harvestervermessung Diameter Error limits Target range Optimisation range Exclusion range Mean [mm] < 1.5 1.5 to 2.5 >2.5 Standard deviation [mm] < 6.0 6.0 to 8.0 > 8.0 Difference of means [%] ± 1.0 Extreme values ( ± 20 mm) < 3,0 % 3.0 % to 5.0 % > 5.0 % Length Mean [mm] < 2.0 2.0 to 3.0 > 3.0 Standard deviation [mm] < 3.0 3.0 to 5.0 > 5.0 Difference of means [%] ± 1.0 Extreme values ( ± 20 mm) < 2.0 % 2.0 % to 5.0 % > 5.0 % 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 15 Material and Methods Necessary action Immediate adjustment Frequent monitoring Differences of diameter (from at least 10 measurements) Mean [mm] Standard deviation [mm] Differences of length (from at least 3 measurements) Mean [mm] Standard deviation [cm] > 2.5 > 8.0 3.0 > 5.0 1.5 6.0 2.0 > 3.0 Source: Dietz & Seeling, 2010 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 16 Source: Dietz & Seeling, 2010 8
Material and Methods PolterLuchs Photo-optical method to automatically measure the number of logs in a round wood pile Vehicle with camera is driving along the log piles and takes series of pictures which iarecompiled by the program (from 3 m distance a 4 m pile can be recorded, at a camera height of 2 m) No spatial reference needed Output: number of logs for each pile Data is compatible with logistic software (e.g. GeoMail) 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 17 Material and Methods Trial set up Harvested stand Norway Spruce Owner: State Forest Nordrhein-Westfalen 50 60 years Max. height 27 m 9 ha 18 % terrain slope Single tree harvest (thinning to favor crop trees) Operating trails at 20 m spacing Harvested volume 50 m³/ha 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 18 9
Material and Methods Trial set up Harvester: John Deere 1470D Eco3-7.72 x 3 m (LxW) - Engine: 552 c.u.-in. 9.0 L, 241 SAE gross hp (180 kw) - Balanced front gear bogie, rigid rear axle - Standard operation weight: 19,700 kg - Reach with harvester head: 10 m Harvester head: Waratah H480C - TimberMatic Software - Weight (without rotator & link) 1240 kg - Cutting Capacity: 650 / 710 mm 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 19 Material and Methods What was done? Harvester was calibrated to standard Spruce stand was thinned, 1345 saw logs were processed Logs were manually measured Logs were extracted Log piles were assessed by conventional manual methods Number of logs in log piles was derived by using PolterLuchs PolterLuchs was used at varying conditions Logs were measured in mill Harvester data was assessed against log identical mill data (diameter, length) and manual measurements 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 20 10
Material and Methods Manual measurements of logs in stand Log assessment at roadside 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 21 Material and Methods Certified laser assessment Diameter, length Taper, ovality 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 22 11
Average log volume [m³ ub] Total volume [m³ ub] 12.03.2014 Results - Harvester 1. Harvester 1.1 Average log volume of harvester compared to mill measurements 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 23 Results - Harvester 1. Harvester 1.1 Average log volume of harvester compared to mill measurements 0,25 Volume comparison [n=1031] 233 242 250 0,20 200 0,15 150 0,10 100 0,05 0,226 0,235 50 0,00 Harvester Mill 0 3.8% difference Average log volume [m³ ub] Total volume [m³ ub] 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 24 12
Number of logs [n] Average log volume [m³ ub] 12.03.2014 Results - Harvester 1.2 Potential factors influencing accuracy of harvester data Number of logs and average log volume per log position 1200 1000 1031 0,35 0,30 800 600 400 200 132 350 338 152 59 0,25 0,20 0,15 0,10 0,05 0 Total bottom log mid log 1 mid log 2 mid log top log 3 0,00 Log position Number of logs [n] Average log volume mill [m³ ub] 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 25 Results - Harvester 1.2 Potential factors influencing accuracy of harvester data 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 26 13
Results - Harvester 1.2 Potential factors influencing accuracy of harvester data 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 27 Results - Harvester 1.2 Potential factors influencing accuracy of harvester data 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 28 14
Results - Harvester 1.2 Potential factors influencing accuracy of harvester data 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 29 Results - Harvester 1.2 Potential factors influencing accuracy of harvester data 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 30 15
bark deduction [cm] 12.03.2014 Results - Harvester 1.2 Potential factors influencing accuracy of harvester data Bark deduction 2,5 2 1,5 1 regional bark deduction table measured deduction 0,5 0 0 1a 1b 2a 2b 3a 3b 4 5 diameter class 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 31 Results - Harvester Findings 1.1 - Average log volume given by harvester is 3.8% smaller than from mill 1.2 - Harvester derived diameter and length data are lower compared to mill assessment except for upper medium and top logs - Log quality (ovality, taper) had no obvious effect on diameter and length accuracy - Bark deduction is dominating effect on diameter for logs with diameter > 29 cm 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 32 16
Polterluchs test scenarios Tested species: Spruce, Douglas Fir, Larch Methods: Section method, End-face method, PolterLuchs PolterLuchs Variation: - brightness setting - discoloration - species - weather (snow) - processing 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 33 Results - Polterluchs 2. PolterLuchs 2.1 Quality of log count by PolterLuchs Amount of correctly recognized logs (Norway Spruce) and time consumption for manual post-processing (clean pile, default setting, no mask) Pile Logs [n] Correctly recognized logs [n] [%] Post-processing time per 100 logs [min] 1 204 186 91% 0.81 2 226 168 74% 0.82 3 579 483 83% 0.67 4 144 105 73% 0.71 5 192 144 75% 1.21 Sum 1345 1086 81% 10.73 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 34 17
Results - Polterluchs 2.2 Potential factors influencing log count accuracy Brightness setting on camera (no mask, clean pile) Pile Logs [n] Correctly recognized logs Post-processing time per 100 logs [min] brightness 30 40 50 60 70 30 40 50 60 70 1 204 185 187 186 182 169 0.82 0.63 0.81 0.67 1.75 2 226 169 168 168 181 181 0.73 0.92 0.82 0.72 0.72 3 579 468 465 483 475 476 0.46 0.67 0.67 0.47 0.58 4 144 100 105 105 108 105 0.66 0.69 0.71 0.82 0.97 5 192 144 141 144 144 149 0.70 0.70 1.21 0.73 0.66 1345 1066 [79%] 1066 [79%] Total 1085 [81%] 1090 [81%] 1080 [80%] Total time for all logs & piles 8.28 9.58 10.73 8.30 11.23 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 35 Results - Polterluchs Brightness setting on camera - Pile 1, brightness 70: 219 out of 204 3.34 min - Pile 1, brightness 30: 189 out of 204 1.42 min 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 36 18
Results - Polterluchs 2.2 Potential factors influencing log count accuracy Discoloration of logs Brightness 70, clean: 504 out of 554 (490 correct), 4.91 min post-processing Brightness 70, discolored: 371 out of 554 (347 correct), 4.96 min post-processing 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 37 Results - Polterluchs 2.2 Potential factors influencing log count accuracy Tree species (no mask, clean pile) Pile Logs [n] Correctly recognized logs 30 50 70 Brightness adjustment at camera Av. Recognized logs Post-processing time per 100 logs [min] 30 50 70 [n] [%] [n] [%] [n] [%] [%] [min] [min] [min] Douglas fir (3 m) 350 211 60 214 61 222 63 61 0.61 0.64 0.97 Douglas fir (4 m) 276 160 58 154 56 195 71 62 1.12 1.29 0.69 Spruce (4 m) 472 398 84 410 87 408 86 86 0.47 0.80 0.39 Larch (3 m) 185 131 71 129 70 128 69 70 0.45 0.54 0.50 Larch (4 m) 103 71 69 55 53 69 67 63 0.70 1.07 0.77 Total Total time for all logs & piles 1385 971 70 961 69 1022 74 71 8.99 11.69 8,85 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 38 19
Results - Polterluchs Douglas fir (3 m) - 223 from 350 (64 %) recognized logs - 2.24 min processing time 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 39 Results - Polterluchs 2.2 Potential factors influencing log count accuracy Weather (without and with snow (50, without mask)) - 235 from 243 recognized logs (of which 231 (95%) were correct) - 1.89 min post-processing time - 136 from 243 recognized logs (of which 120 (49%) are correct) - 4.55 min post-processing time 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 40 20
Results - Polterluchs 2.2 Potential factors influencing log count accuracy Processing with and without mask 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 41 Results - Polterluchs Processing without mask with mask 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 42 21
Results - Polterluchs 2.2 Potential factors influencing log count accuracy Processing with and without mask Pile Logs [n] Correctly recognized logs Brightness adjustment at camera Post-processing time per 100 logs [min] 50 (no mask) 50 (mask) 50 (no mask) 50 (mask) [n] [%] [n] [%] [min] [min] Douglas fir (3 m) 350 214 61 214 61 0.64 1.00 Douglas fir (4 m) 276 154 56 156 57 1.29 1.93 Spruce (4 m) 472 410 87 413 88 0.80 0.86 Larch (3 m) 185 129 70 129 70 0.54 0.78 Larch (4 m) 103 55 53 58 57 1.07 2.53 Total Total time for all logs & piles 1385 961 69 970 74 11.69 15.59 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 43 Results - Harvester Findings 2.1 - In average 81% of the Spruce logs were recognized 2.2 - Brightness setting is of particular importance under light conditions - Discoloration of logs did not lead to increased processing effort - Snow cover of pile reduced recognition rate and increased processing time - Douglas Fir and Larch showed lower recognition rates and increased processing time - Processing with mask did not shorten overall processing time 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 44 22
Diameter Length 12.03.2014 Overall Results Achieved accuracies compared to standard Calibration settings KWF Error limits Target area Optimization area Exclusion area Mill vs. Harvester (n=1304) Mean [mm] < 1,5 1,5 x 2,5 > 2,5 3,45 Standard deviation [mm] < 6,0 6,0 x 8,0 > 8,0 12,38 Difference of mean [%] ± 1,0 1,4 Extreme values ( ± 20mm) [%] < 3,0 3,0 x 5,0 > 5,0 12,4 Mean [cm] < 2,0 2,0 x 3,0 > 3,0-0,1 Standard deviation [cm] < 3,0 3,0 x 5,0 > 5,0 4,74 Difference of mean [%] ± 1,0 1 Extreme values ( ± 10 cm) [%] < 2,0 2,0 x 5,0 > 5,0 1,4 Δ Volume [%] 3,6 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 45 Overall Results Accuracy and time consumption of pile assessment methods Scenario Scaling method Volume [m³sub] Volume in relation to mill volume [%] Time consumption [min] Time consumption [min/ m³sub] Time consumption in relation to end-face method [%] Status quo 1 sections 303.69 96.9 % 170.40 0.54 84.4% Status quo 2 end-face 335.63 107.1 % 200.37 0.64 100.0% Status quo 3 manual, logwise Mill automized, certified Harvester harvester head Combination Harvester + of avg. Log polterluchs volume x number of logs 325.82 104.0 % 313.44 100.0 % 302.40 96.5 % 302.40 96.5 % 43.46 0.14 21.8% 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 46 23
Results Questions of interest: 1. Harvester 1.1 How accurate is the average log volume derived by a harvester calibrated to standard compared to mill measurements? 1.2 What are potential factors influencing accuracy of diameter and/or length measurements of the harvester 2. PolterLuchs 2.1 How accurate is the log number of log piles assessed by this method 2.2 What are potential factors influencing accuracy of derived log numbers 3. How accurate are derived harvest volumes by harvester/polterluchs compared to manual and mill assessments of volume 05.03.2014 Dirk Jaeger Integration of high quality data and new allocation technology for efficient control of wood flow 47 Thank you for your attention! Special thanks and appreciation goes to Mr. Thilo Wagner, Director of the forest education center in Arnsberg, for his continuous support. Prof. Dr. Dirk Jaeger University of Freiburg Chair of Forest Operations phone +49 761 203 35 67 e-mail dirk.jaeger@fobawi.uni-freiburg.de 24