On the Methods to Predict Moisture Content on Wood: A Literature

Review

ıtor M. Magalh

aes

, Giancarlo Lucca

, Alessandro De L. Bicho

and Eduardo N. Borges

Centro de Ci

encias Computacionais, Universidade Federal do Rio Grande – FURG, Brazil

Keywords:

Moisture Content, Wood, Intelligent Systems, Machine Learning, Prediction.

Abstract:

Wood is the raw material for many manufactured goods. Charcoal, cellulose for the paper industry, laminated

wood furniture, and even explosive products, such as gunpowder cotton, are possible destinations for the wood.

On the other hand, the growing use of wood as a raw material has increased illegal deforestation and, as a direct

consequence, it has changed the climate at a global level. The use of wood in production processes must be

optimized to mitigate these adverse effects. One of the determining factors for this optimization is moisture

content on wood, i.e., the ratio between the mass of water contained in the wood and dry wood mass. This

article reviews the scientiﬁc literature published from 1959 to 2019 regarding the use of wood due to a better

knowledge of its properties, particularly systems to explain or predict the moisture content. It contributes to

the continuity of related research with the theme by ensemble the conducted studies into a single analysis.

1 INTRODUCTION

China’s exponential growth and its consequent struc-

tural transformation from an essentially rural society

to an urban-industrial civilization would inevitably

produce demands for natural resources of all kinds.

The policy of China’s economy has brought about a

dramatic expansion in foreign trade (Hu and Khan,

1997). Over the past twenty years, Brazil has be-

come the largest supplier of agricultural products to

the Chinese market. Exports to China accelerated and

deepened technological and organizational changes in

Brazilian agriculture – the base of the competitiveness

of Brazilian agribusiness (Vieira et al., 2019).

Considering the wood, given its limited access to

natural resources, China has increasingly resorted to

external purchases for the production of its local fac-

tories. This fact led Brazil to double its exports of

wood logs between 2017 and 2018. According to a

survey by the Forest2Market consultancy, based on

the Brazilian government’s foreign trade data, exports

of eucalyptus logs in the year 2018 increased 122%

compared to 2017. Of this total, 89% was exported to

China (Brazilian Society of Agriculture, 2019).

https://orcid.org/0000-0003-3588-9930

https://orcid.org/0000-0002-3776-0260

https://orcid.org/0000-0002-6572-1496

https://orcid.org/0000-0003-1595-7676

It is also necessary to add the Brazilian national

demand, already have been analyzed for the medium

and long term. Precisely, between 1988 and 2007,

studies already presented results indicating that the

variables price of wood logs and installed capacity

of the pulp industry explained the Brazilian national

demand (

Angelo et al., 2009).

Almost all ﬁnished products with wood as raw

material go through the same stage: storing wood logs

in piles. To better understand this question, Figure 1

shows a pile of wood logs in a storage yard.

The condition of the wood logs during storage is a

determining factor for the quality of the ﬁnished prod-

uct. The storage time of wood logs directly inﬂuences

acceptance, if the wood is destined for the manufac-

Figure 1: Example of wood logs stored in a pile.

Magalhães, V., Lucca, G., Bicho, A. and Borges, E.

On the Methods to Predict Moisture Content on Wood: A Literature Review.

DOI: 10.5220/0011063100003179

In Proceedings of the 24th International Conference on Enterprise Information Systems (ICEIS 2022) - Volume 1, pages 521-528

ISBN: 978-989-758-569-2; ISSN: 2184-4992

521

ture of wood chips (Lima et al., 2017) or laminates of

wood (J

unior and Alves, 2019), two of the essential

purposes of wood in the Asian market.

Logs of wood from newly felled trees have high

water content. This content is slowly reduced as the

logs are exposed to the environmental conditions dur-

ing storage (Rezende et al., 2010). The wood starts

to dry, and its weight decrease with the loss of mois-

ture (Tomczak et al., 2018).

Wood can absorb from 25 up to 30% of its weight

in water (Dorigato et al., 2019). As the wood dries,

cracks start to occur. The cracks are responsible

for the large volume of wood losses during log stor-

age (Gerwing et al., 2000). Thus, the moisture con-

tent must be observed and kept within the standards

during the storage time (Estuqui Filho, 2006). Figure

2 shows examples of cross-section images of wood

logs, stored in piles, with different crack conditions.

Figure 2: Cracks in wood logs.

There is a direct relationship between the mois-

ture content and the weight of wood logs stored in the

yards. Based on this premise, this work aims to re-

search the relevant scientiﬁc literature that addresses

the problem of modeling the variation in the mois-

ture content and weight of logs. This research can be

used for building different intelligent information sys-

tems, helping to create new technologies, especially

that which use machine learning methods.

Machine Learning is a subarea of Artiﬁcial Intel-

ligence (AI) (Russell and Norvig, 2002) and can be

considered as an evolving branch of computational

algorithms designed to emulate human intelligence,

learning from the environment (El Naqa and Murphy,

2015). Machine learning allows the computer to de-

velop models that automatically learn about a partic-

ular topic. Machine learning-based techniques have

been successfully applied in many ﬁelds, from pattern

recognition, computer vision, spacecraft engineering,

centralized and decentralized ﬁnance, entertainment,

computational biology, and medical applications.

This paper is organized as follows. We ana-

lyze two different ways to deal with the variation

of wood moisture content: using machine learning-

based models (Section 2) and analytical and statistical

approaches (Section 3). We still present in Section 4

other methods that indirectly contribute to solving this

problem. We deeply discuss some interesting studies

for each method. In Section 5, we organize, summa-

rize and categorize the literature reviewed in a time-

line format, demonstrating the evolution of related re-

search. Finally, Section 6 concludes our study and

points out some future directions.

2 MACHINE LEARNING-BASED

MODELS

Machine Learning (ML) is a computer program that

learns from experience concerning some class of

tasks, evaluating its performance when executing

these tasks improves with the experience (Mitchell,

1997). ML can be understood as a subset of Arti-

ﬁcial Intelligence (AI) that has the ability to build

mathematical models for the purpose of making pre-

dictions or decisions without having to be explicitly

programmed to do this (Zhang, 2020).

A moisture content modeling method, was pro-

posed by using Support Vector Machine (SVM) tech-

nique (Wen et al., 2012). In this model, the input

is the temperature and the equilibrium moisture con-

tent, i.e., the moisture content at which wood neither

gains nor loses moisture. The moisture content is the

output. The training data was obtained from a wood

drying kiln with a temperature between 26.6 and 37.7

ºC and steam heating. Besides using the drying kiln,

there were still heating equipment and measuring in-

struments, including temperature sensors, moisture

content sensors, equilibrium moisture content sen-

sors, drought fans, and controllers. Moreover, exper-

imental parameters selection for the SVM model was

investigated, and the inﬁnite impulse response (IIR)

ﬁlter technique was applied to screen noise of the

training data. The results show the effectiveness of

the proposed method, demonstrating that it is possible

to use machine learning techniques for this purpose.

Fuzzy logic was also applied to predict moisture

content (Bardak and Bardak, 2019). This work was

related to the drying temperature and time. The au-

thors have used the An air-dried Fagus orientalis

species, which is commonly wood utilized in man-

ufacturing. The specimens were dimensioned as

55×25×25mm. Then, the samples were kept in the

moisture-conditioned room until weight gain reached

equilibrium. The moisture content of 13% was

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

522

achieved. Afterward, wood samples were dried in

a drying oven at different temperatures (50ºC, 70ºC,

90ºC, 110ºC) and time (0.5, 1.0, 1.5, 2.0, 2.5 hours).

Ten samples were prepared for each experiment. This

model has two input variables (temperature, time)

and one output variable (moisture content). The ﬁt-

ted model showed an average accuracy rate equal to

97.16%, showing that fuzzy logic can be used as a

valuable tool in the wood drying process, which is an

important part of the cost in the wood industry.

Machine learning algorithms were also used to

predict wood volumes in trees and better allocation

of wood logs for different uses, comparing them with

the taper equations (Souza, 2019). These equations

explain the relationship between the diameter and the

height of a tree. Unlike traditional approaches in

which wood volume is based on a single measure of

tree diameter, taper equations use trunk taper mea-

surements, providing considerably more accurate vol-

ume estimates, considering the changes in decreasing

diameters from the base to the top of the trees. The

authors compared Artiﬁcial Neural Networks (ANN),

Support Vector Regression (SVR), and Random For-

est (RF), trying to understand the behavior of the pre-

dictions proposed by different algorithms, in addition

to specifying the best models for the studied cases.

The ANN model reached the best accuracy. RF gener-

ated imprecise estimates unless the measurements are

taken at smaller intervals and with large amplitudes of

log size classes. The ML algorithms performed better

or equal to taper equations.

To organize wood logs used as raw material in

sawmills, the authors (Morin et al., 2020) demon-

strate how machine learning models can provide rec-

ommendations for multiple tasks: allocation of cut-

ting blocks, helping in the planning stage, and rec-

ommending both the wood and the mill that will pro-

cess it. In this way, they exploit the processing plants’

strengths and reduce losses in the process. The pro-

posed models achieved up to 94% of the maximum

theoretical gain in decision-making using Decision

Trees (DT), with an average gain of 83% compared

to the more usual method of historical production.

Regarding the variation in moisture content, we

highlight a study that used ANN to predict the drying

rates of wood kiln based on species and basic wood

density information (Wu and Avramidis, 2006). Al-

though this is not precisely a sub-area of interest in

this research because of kiln-drying rather than air-

drying, the contribution to estimating wood moisture

content using machine learning must be considered.

The neural network developed had three inputs: ini-

tial moisture content, density, and drying time. The

model’s output was the estimate of the average ﬁ-

nal moisture content. A back-propagation algorithm

was implemented to train, validate, and test the model

with 50, 25, and 25% of instances, respectively. The

formation of these partitions was due to the equal

spacing of the points and the original data. After

obtaining the optimized conﬁguration by varying the

main hyperparameters, such as the transfer function,

the learning rate, and the number of neurons and

learning layers, the quality of the ﬁnal model achieved

a Mean Relative Absolute Error (MRAE) lower than

2%, conﬁrming the excellent predictive capacity of

this method in order to predict the drying rate.

3 ANALYTICAL, STATISTICAL,

OR GENERALIZATION-BASED

MODELS

In Finland, researchers estimated the ideal storage

time of woodpiles kept outdoors based on moisture

content changes (Raitila et al., 2015). Multivariate

models were created aiming to estimate moisture con-

tent changes in different drying environments, based

on testing and validating models developed for the pi-

nus wood species. These models were applied to piles

of wild pinus species destined for fuel production. In

order to cover different moisture levels through the

different periods of the year, the experimental data

were collected over seven to fourteen months and

considered the precipitation, evaporation, and wood

species as main factors. The work compared regres-

sion models to estimate moisture content with the

more traditional method – the periodic weight dif-

ference of logs stored in piles. The conclusion was

that data logging based on load cells to estimate the

change in wood moisture content proved to be a vi-

able option to obtain the necessary data for wood dry-

ing models.

Regarding the wood drying process, there are re-

searches directly related to the construction of met-

rics for different estimates. The theme has been re-

searched since 1959, and it is known that it is ruled

by some variables, such as temperature, relative hu-

midity, wind speed, and precipitation (Kr

oll, 1959).

In 2009, another study inserted more variables to be

studied and researched, such as access to wind cur-

rents and exposure to the sun (Kofman and Kent,

2009) in the outdoor drying process (air drying).

In 2000, a computer simulation was developed an

analytical way of estimating the natural drying times

of various species of laminated wood from local mete-

orological data for wood piled on any day of the year

(Simpson, 2000). In this study, the effect of sawn

On the Methods to Predict Moisture Content on Wood: A Literature Review

523

wood thickness on drying time was also included.

The model presented is based on experimental times

of natural drying for six wood species: northern red

oak, sugar maple, American beech, yellow poplar,

pinus ponderosa and Douglas-ﬁr. The premise was

that, once the parameters were found for each wood

species in the geographic location of the experimental

data, they could be used in drying simulation, estimat-

ing natural drying times in other locations where his-

torical meteorological data were also available. The

results show the estimated natural drying times of the

six wood species, for any thickness, up to any ﬁnal

moisture content, on any day of the year and in any

location where there are data on average temperature

and relative humidity.

In 2004, a study by the same researcher (Simpson,

2004) developed a non-linear regression model to de-

scribe outdoor drying equally for two wood species.

In this research, regression coefﬁcients were devel-

oped to predict the daily loss of moisture content

based on the same variables: early-day moisture con-

tent, temperature, and average daily relative humidity.

The wood species used in this study were northern red

oak and sugar maple. The resulting regression mod-

els are presented below:

= MC

2.38

0.759

−2.91

(northern red oak)

= MC

2.18

1.89

−3.57

(sugar maple)

The models can predict daily moisture loss based on

almost always the same variables (MC is early wood

moisture content, F is daily average temperature, and

RH is daily average relative humidity). Therefore,

having the variables in common, the equations could

be used to develop curves of loss of moisture content

over time, as long as local meteorological data were

available, observing only the wood species. So, the

research resulted in the development of an electronic

spreadsheet to estimate the time needed for drying

wood outdoors, based on an index called Air-Drying

Estimator (Mitchell, 2019).

Corroborating the above studies, another study

demonstrates that meteorological data used in natu-

ral drying models can adequately estimate the wood

moisture content in logs (Erber et al., 2012). Also,

considering that determining the moisture content in

a log pile without measuring it frequently is an opera-

tional problem, this study aimed to model the change

in the moisture content of a log pile as a function of

measurable input variables. So, the periods of natu-

ral drying could be optimized, and thus, the period

of transformation of wood into ﬁrewood (for energy

production) could be determined.

In order to estimate these data, equipment was

used to constantly measure - every ten minutes -

the wood samples in logs. The data collected were:

percentage of relative humidity, temperature, wind

speed, and precipitation. For calculation purposes,

even if it was in the form of snow - non-liquid pre-

cipitation, the precipitation was transformed into the

equivalent of net precipitation. The regression model

then generated was:

CMC = 4.691 × 10

−3

DP + 1.359 × 10

−2

where CMC is the daily percentage of moisture con-

tent variation, T is the temperature (measured in de-

grees Celsius), DP is the percentage of potential daily

relative humidity, and P is the sum, in millimeters, of

the daily net precipitation and non-liquid.

The objective of this multiple regression model

was to be based on daily averages and sums, seek-

ing to operationalize models previously studied more

easily and quickly, not requiring data from the logs

themselves. An important piece of information raised

by this study concerns wood species. The ﬁndings

on precipitation data do not match with other pre-

vious studies, which claimed that precipitation had

no signiﬁcant effect on natural drying (Gigler et al.,

2000). Thus, it is assumed that certain wood species,

such as wild pinus and willow have only surface mois-

ture. Furthermore, wild pinus can be considered more

likely to retain moisture in the bark than willow.

4 OTHER RELATED WORKS

On the storage of wood logs, a study investigated the

effect caused by the storage time of wood logs in New

Zealand (Visser et al., 2014). In this work, two tri-

als were performed representing favorable and unfa-

vorable storage conditions: in summer, in a hot and

dry location, and in winter, in a cold and relatively

humid place. Twenty piles were installed, contain-

ing approximately 600 kg of wood logs each (ini-

tial wet weight). Moisture content was determined

gravimetrically at the tests’ beginning and end. All

piles were weighed at 1 to 4-week intervals to identify

weight loss trends over storage time. After twenty-

four weeks of storage in the summer, the wood logs’

moisture content (wet base) decreased from an initial

value of 53% to values between 33 and 21%. The

decrease was more signiﬁcant for small uncovered

logs and smaller for large covered logs. Due to the

wet and cold weather conditions, logs stored in win-

ter dried very little over seventeen weeks. Moisture

content decreased from an initial value of 58% to val-

ues between 51 and 49%, with no signiﬁcant treat-

ment differences observed in the winter test. The best

storage technique for the summer was the simplest:

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

524

stacking small logs without any covering. Larger logs

dried more slowly but splitting accelerated their dry-

ing. Stack covering did not help decrease moisture

loss, and the results indicated that log covering did

not improve the drying of wet logs.

Wood can be stored in piles for a variety of pur-

poses. One of them is for energy production. A study

produced in Sweden (Th

ornqvist, 1985) analyzed the

relationship between moisture loss and loss in energy

production. In order to measure and determine the

changes in energy production due to storage, moisture

and matter loss (weight loss) before and after storage

were analyzed. This study concluded that if the wood

stored in piles is reduced to chips instead of wood

logs, the energy loss can vary between 7% and 21%

for a period between six and nine months of storage.

It is known that recently felled wood has high wa-

ter content and that water loss is a preponderant fac-

tor through the loss of moisture content. The study

(Tomczak et al., 2018) aims to demonstrate the min-

imum period for wood logs stored in piles to start

the natural drying process. To determine how the

method and storage conditions affect weight changes

and moisture loss in the wood, a plot of sixty model

trees was selected, divided into two groups: in the ﬁrst

group, thirty whole trees; in the second group, thirty

trees were cut transversely, forming logs with 2.5 m

in length and stored in piles. From experimentation

in this study, it was noted that wood stored in piles

lost moisture more slowly than wood from trees that

were left whole after felling. Comparing the weights

of logs stored in a pile days after harvesting, a statis-

tically signiﬁcant difference was found only between

the ﬁrst and ﬁfteenth – and last – day. Therefore, this

study concluded that the two-week period is the min-

imum pile storage period necessary to achieve a sig-

niﬁcant degree of weight change and moisture loss.

When it comes to the production of sawn wood

and wood veneer, an study investigated the poten-

tial use of different eucalyptus species (benthamii,

deanei, E. dorrigoensis, E. dunni and E. smithii) in

a region with the occurrence of frost (Walker, 2006).

The trees selected for sampling were eighteen years

old and were sectioned into logs of 2.1 m in length.

Afterward, the samples were subdivided into classes

by diameter: between 20.1 cm and 25 cm; between

25.1 cm and 30 cm; and larger than 30 cm. The cracks

in the boards were measured, and then the yield on al-

ready sawn wood could be calculated. The results ob-

tained in this study indicated that the wood species

studied are of good quality, highlighting E. dunnii.

However, the other species also showed satisfactory

yields, indicating that these can be used to produce

sawn wood in regions with frost occurrence. For the

present research, an outstanding contribution of this

work is the relationship of cracks with the yield of the

wood to be laminated. This yield is expressed as a

percentage obtained by the ratio between the volume

of sawn wood and the volume of logs exposed to the

process. For eucalyptus, this percentage varies be-

tween 40 and 65%. Among some factors that have a

direct inﬂuence on wood yield, in addition to the char-

acteristics of the species itself, there is also the qual-

ity and diameter of the logs to be processed (Tsoumis

et al., 1991). For this reason, the division of logs into

classes by diameter is an important strategy for the

production of laminated wood with high yield.

5 A COMPARISON AMONG THE

DIFFERENT APPROACHES

This section aims to compare the different approaches

reviewed in this paper. Considering that we provide

as much important information per study as we could

(nineteen different studies), in order to ease the com-

prehension of the summary, we have divided the anal-

ysis into two parts, Tables 1 and 2. In each table, we

provide different studies, one per row and their dif-

ferent characteristics per column. We highlight that

the ﬁrst column, ID, is related to identifying the con-

sidered study. That is, the ID that appears in Table 1

refers to the same study in Table 2.

In Table 1, we provide the year of the publication,

the kind of wood related to the study, the species, the

country related to the study, and the objective of each

study related to moisture content. Also, considering

the published studies, it is noticeable that this topic

has gained attention in the last years, being a problem

with an interesting research ﬁeld. We can also observe

that North America provides most of the studies, and

the United States has around 30% of the analysis (6

cases), followed by Canada (3 cases). We highlight

the different studies that were carried in Europe.

Concerning the motivations, we can observe that

this is an extensive research ﬁeld that has been coped

with machine learning and mathematical problems.

The objectives are economical, optimization, and

study the variation in moisture content, especially air

drying-based models.

The second part of the analysis is provided in Ta-

ble 2. In this table, we provide the adopted method

and the main contribution for each considered study.

We can observe that the question related to machine

learning and the weight of the logs are the methods

most commonly used because there is a direct rela-

tionship between these two critical variables.

On the Methods to Predict Moisture Content on Wood: A Literature Review

525

Table 1: Timeline and researches characteristics.

ID Year Kind of

wood

Species Country Objective

1 1959 Logs (not speciﬁed) Germany To study the wood drying process

2 1982 Laminated Red Oak and Yellow Poplar United States To estimate the drying time

3 1985 Logs /

Chips

(not speciﬁed) Sweden Analyze the relationship between mois-

ture loss and energy production

4 2000 Laminated Northern Red Oak, Sugar

Maple, American Beech,

Yellow Poplar, Pinus Pon-

derosa and Douglas-Fir

United States To estimate the air drying time

5 2003 Logs Pinus Ponderosa and

Douglas-Fir

United States Estimate moisture content loss through

natural drying on small diameter logs

6 2004 Laminated Northern Red Oak, Sugar

Maple and Pinus Ponderosa

United States Describe, through an equation, the air

drying

7 2006 Laminated E. Benthamii, E. Deanei, E.

Dorrigoensis, E. Dunni and

E. Smithii

New Zealand The potential use of different eucalyp-

tus species based on the cracks

8 2007 Logs Giant-Fir and Tsuga Canada To predict drying rates for wood kilns

9 2009 Logs (not speciﬁed) Ireland Estimate the rate of moisture loss in

logs in air drying

10 2012 Logs Pinus Finland Modeling the change in the moisture

content of a log pile as a function of

measurable input variables

11 2012 Logs Pinus Taeda, Eucalyptus

Dunni and Eucalyptus

Grandis

Brazil and

United States

Examine the impact of time on book in-

come and the operating cost of losing

weight

12 2012 (not speci-

ﬁed)

Sugi Japan To predict the moisture content

13 2014 Logs Pinus Radiata New Zealand Check the effect caused by the storage

time of the wood logs

14 2015 Firewood Pinus and Wild Pinus Finland Estimate the optimal storage time for

wood piles stored outdoors based on

changes in moisture content

15 2017 Wood chips Pinus and Oak France To predict the moisture content

16 2018 Logs Wild Pinus Poland Evaluate changes in mass and moisture

content of wood stored in logs

17 2018 Logs Beech Poland Determine the minimum period for logs

stored in piles to start the natural drying

process

18 2019 Wood chips Fagus orientalis Turkey To predict the moisture content

19 2019 Logs /

Laminated

(not speciﬁed) Canada Allocate wood cutting blocks in order

to optimize sawmill production

20 2019 Logs

(trees)

Eucalyptus Grandis Brazil Estimate wood volume in trees

21 2019 Logs (not speciﬁed) Canada Economic waste related to wood weight

loss through moisture loss by natural

drying

22 2019 Laminated Red Oak, Yellow Poplar

and Sugar Maple

United States Estimate the potential of a geographic

location to dry wood

6 CONCLUSION

Wood science and technology in China is becoming

a solid scientiﬁc discipline and has impacted global

research and development activities. The wood dry-

ing process has a complex and non-linear system with

time-varying and uncertain characteristics. Wood

drying is a process of removing water, that is, the de-

creasing process of moisture content. This is a critical

issue in applications where wood and the quality are

affected directly by the change of moisture content.

The aim of this work was to survey the bibliogra-

phy about the properties for better use of wood in pro-

duction processes, precisely the relation between pro-

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

526

Table 2: Timeline and researches contributions.

ID Method Contribution

1 Studied wood drying in drying kilns Established temperature, relative humidity, wind speed

and precipitation as variables

2 Deducing daily rate of moisture loss from meteorologi-

cal data

Deﬁned a regression to estimate the daily loss of mois-

ture content

3 Analyzing moisture and weight loss before and after

storage

Concluded that if the wood stored in piles is reduced to

chips, the energy loss can vary between 7% and 21%

for a period between 6 and 9 months of storage

4 A computer simulation, including the effect of wood

thickness on drying time

Estimated the air drying time, up to any ﬁnal moisture

content, on any day and in any location where there are

the average temperature and relative humidity data

5 Using weather data, measuring temperature and relative

humidity every 10 minutes

Two different equations, one for each species

6 Using historical records of temperature and relative hu-

midity

Developed an equation that allows the prediction of es-

timated air drying time at any desired location

7 Measuring the cracks and calculating the yield on al-

ready sawn wood

Expressed the relationship between cracks and lami-

nated wood yield and established the division of logs

into classes as an important strategy

8 Machine Learning using three inputs: initial moisture

content, density and drying time

Conﬁrmed the excellent predictive capacity of this mod-

eling method

9 Continuously weighing the logs Established access to wind currents and exposure to the

sun as determinants for the estimation of moisture

10 Measure the percentage of relative humidity, tempera-

ture, wind speed and precipitation in log wood samples

Established an equation

11 Constantly weighing Proved that the difference between the green wood and

logs with 11 weeks of cut can increase a lot the cost for

the companies

12 Machine learning method is proposed by using SVM The results of this paper show the effectiveness of the

proposed method, demonstrating that it is possible to

use machine learning techniques for this purpose.

13 Weighed the logs every 1 and 4 weeks The best storage technique for the summer was just

stacking small logs without any covering. Also, it con-

cluded that larger logs dried more slowly

14 Compared regression models (estimating moisture loss)

with a classic model, providing the weight of log piles

Proved that periodic weighing of wood logs is an effec-

tive method to predict the difference in moisture content

15 Continuously weighing the logs Established that as long as is the length of logs stored in

the piles, lower is the loss of moisture

16 Comparing log weights with whole trees (not cutted into

logs)

Concluded that two weeks is the minimum period nec-

essary to achieve a signiﬁcant moisture loss

17 Relate machine learning models with a existing simula-

tor

Machine Learning models reached up to 94% of the

maximum theoretical gain

18 Machine learning using as input data provided by the

coefﬁcient of reﬂection

The result show the effectiveness of the proposed mod-

eling methodology, allowing this solution for moisture

content prediction to be suitable for direct implementa-

tion on real-time wood-to-energy industrial processes.

19 Compared taper equations with models based on Ma-

chine Learning

Methods using Machine Learning were equal to or su-

perior to the method using taper equations

20 Continuously weighing the logs A mathematical formulation is proposed that decides on

the location of potential wood storage yards

21 Fuzzy logic, relating the moisture content with the dry-

ing temperature and time

The results of the fuzzy models showed an average ac-

curacy rate of 97.16%, showing that it can be used as a

useful tool in the wood drying process – an important

part of the cost in the wood industry

22 Simulations Outputs: the estimated date when the wood will reach

the target moisture content and others

ductivity and moisture content. By carrying out this

study, it will be possible to help different information

systems as the data are organized and summarized.

It is also known that wood properties can vary

widely, depending on storage and climatic conditions

and the destination of the wood itself. Therefore,

On the Methods to Predict Moisture Content on Wood: A Literature Review

527

an understanding of these properties, especially those

that are technically important to measure losses - such

as the behavior of moisture content - can increase the

potential for optimizing the use of wood, regardless

of whether the destination is lamination, the chip pro-

duction or even power generation. Increasing the po-

tential for using the same wood will inevitably help

mitigate the acceleration of climate change globally.

ACKNOWLEDGEMENTS

This study was supported by CNPq (305805/2021-5)

and PNPD/CAPES (464880/2019-00).

REFERENCES

Angelo, H., Almeida, A. N. d., and Serrano, A. L. M.

(2009). Determinantes da demanda de madeira em

toras para celulose no brasil. Scientia Forestalis,

37(84):491–498.

Bardak, S. and Bardak, T. (2019). Moisture content predic-

tion of wood drying process using fuzzy logic. Acta

Scientiarum Polonorum Silvarum Colendarum Ratio

et Industria Lignaria, 18(1):71–75.

Brazilian Society of Agriculture (2019). Export of

eucalyptus logs doubles with chinese demand (in

portuguese). https://www.sna.agr.br/exportacao-de-

tora-de-eucalipto-dobra-com-demanda-chinesa/. Ac-

cessed: dec., 2021.

Dorigato, A., Fredi, G., and Pegoretti, A. (2019).

Thermo-mechanical behavior of novel wood laminae-

thermoplastic starch biodegradable composites with

thermal energy storage/release capability. Frontiers

in Materials, 6. Article 76.

El Naqa, I. and Murphy, M. J. (2015). What Is Machine

Learning?, pages 3–11. Springer Intl. Pub., Cham.

Erber, G., Kanzian, C., Stampfer, K., et al. (2012). Predict-

ing moisture content in a pine logwood pile for energy

purposes. Silva Fennica, 46(4):555–567.

Estuqui Filho, C. A. (2006). A durabilidade da madeira

na arquitetura sob a ac¸

ao dos fatores naturais: estudo

de casos em bras

ılia. Master’s thesis, Faculdade de

Arquitetura e Urbanismo - Universidade de Bras

ılia.

Gerwing, J., Vidal, E., Ver

ıssimo, A., and Uhl, C. (2000).

Rendimento no processamento de madeira no estado

do Par

a. S

erie Amaz

onia, 18:38.

Gigler, J. K., van Loon, W. K., Van den Berg, J., Sonneveld,

C., and Meerdink, G. (2000). Natural wind drying of

willow stems. Biomass & Bioenergy, 19(3):153–163.

Hu, Z. F. and Khan, M. S. (1997). Why is china growing so

fast? Staff Papers, 44(1):103–131.

unior, A. S. and Alves, J. E. (2019). Fatores intervenientes

no armazenamento de laminados. Gepros: Gest

ao da

Produc¸

ao, Operac¸

oes e Sistemas, 14(5):190.

Kofman, P. D. and Kent, T. (2009). Long term storage and

seasoning of conifer energy wood. Coford connects–

Harvesting/Transportaions, 20:1–4.

oll, K. (1959). Selbstt

atiges steuern und regeln der trock-

ner. In Trockner und Trocknungsverfahren, pages

548–563. Springer.

Lima, R., da Silva Cardoso, G., de Proenc¸a, G., and

da Costa, W. G. (2017). Inﬂu

encia do tempo de ar-

mazenamento (tpc) da madeira no aceite de cavacos

de eucalipto para a fabricac¸

ao de polpa. In Congresso

Internacional de Celulose e Papel. ABTCP.

Mitchell, P. H. (2019). Development of an air drying index

for lumber. BioResources, 14(3):6657–6665.

Mitchell, T. (1997). Machine learning. MacGraw-Hill.

Morin, M., Gaudreault, J., Brotherton, E., Paradis, F., Rol-

land, A., Wery, J., and Laviolette, F. (2020). Machine

learning-based models of sawmills for better wood al-

location planning. Int. Journal of Prod. Econ., 222.

Raitila, J., Heiskanen, V.-P., Routa, J., Kolstr

om, M., and

Sikanen, L. (2015). Comparison of moisture predic-

tion models for stacked fuelwood. BioEnergy Re-

search, 8(4):1896–1905.

Rezende, R. N., Lima, J. T., de Ramos, L. E., Faria, A.

L. R., et al. (2010). Secagem ao ar livre de toras de

eucalyptus grandis em Lavras/MG. Cerne, 16:41–47.

Russell, S. and Norvig, P. (2002). Artiﬁcial intelligence: a

modern approach. Prentice Hall.

Simpson, W. T. (2000). Estimates of air drying times for

several hardwoods and softwoods, volume 121. US

Dept. of Agriculture, Forest Service, Forest Prod. Lab.

Simpson, W. T. (2004). Estimating air drying times of lum-

ber with multiple regression, volume 293. US Dept of

Agriculture, Forest Service, Forest Prod. Lab.

Souza, G. S. A. d. (2019). Aprendizado de m

aquina em

aplicac¸

oes de manejo ﬂorestal. PhD thesis, UFV.

ornqvist, T. (1985). Drying and storage of forest residues

for energy production. Biomass, 7(2):125–134.

Tomczak, A., Grodzi

nski, G., Jakubowski, M., Jelonek, T.,

and Grzywi

nski, W. (2018). Effects of short-term stor-

age method on moisture loss and weight change in

beech timber. Croatian J. of Forest Eng, 39(1):35–43.

Tsoumis, G. et al. (1991). Science and technology of wood:

structure, properties, utilization, volume 115. Van

Nostrand Reinhold New York.

Vieira, P. A., Buainain, A. M., and Figueiredo, E. V. C.

(2019). O brasil alimentar

a a china ou a china engolir

o brasil? Tempo do Mundo, 2(1):51–82.

Visser, R., Berkett, H., and Spinelli, R. (2014). Determining

the effect of storage conditions on the natural drying

of radiata pine logs for energy use. New Zealand Jour-

nal of Forestry Science, 44(1):1–8.

Walker, J. C. (2006). Primary wood processing: principles

and practice. Springer Science & Business Media.

Wen, S., Deng, M., and Inoue, A. (2012). Moisture con-

tent prediction of wood drying process using svm-

based model. Int. J. Inov. Comput. Infomation Con-

trol, 8:4083–4093.

Wu, H. and Avramidis, S. (2006). Prediction of timber kiln

drying rates by neural networks. Drying Technology,

24(12):1541–1545.

Zhang, X.-D. (2020). Machine learning. In A Matrix Al-

gebra Approach to Artiﬁcial Intelligence, pages 223–

440. Springer.

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

528