NC State
BioResources
Borz, S. A., Talagai, N., Cheţa, M., Gavilanes Montoya, A. V., and Castillo Vizuete, D. D. (2018). "Automating data collection in motor-manual time and motion studies implemented in a willow short rotation coppice," BioRes. 13(2), 3236-3249.

Abstract

Time and motion studies are often used to evaluate the performance of various product systems. However, traditional studies are characterized by a series of technical limitations, and they require many resources. This study tested the capability of a low-cost Global Positioning System (GPS) receiver and an accelerometer unit to automate the field data collection for characterizing motor-manual felling of willow short rotation coppices. The results were promising. By thresholding the acceleration data, the running and stopped engine states were accurately separated. Also, by combining the GPS speed with the acceleration data, followed by threshold setting and data visualization in the Geographic Information System software, detailed time categories, such as productive, working, and non-working times, could be separated. The methods described herein could be used to manage long-term field data collection, as such operations are affected by many operational factors.


Download PDF

Full Article

Automating Data Collection in Motor-manual Time and Motion Studies Implemented in a Willow Short Rotation Coppice

Stelian A. Borz,* Nicolae Talagai, Marius Cheţa, Alex V. Gavilanes Montoya, and Danny D. Castillo Vizuete

Time and motion studies are often used to evaluate the performance of various product systems. However, traditional studies are characterized by a series of technical limitations, and they require many resources. This study tested the capability of a low-cost Global Positioning System (GPS) receiver and an accelerometer unit to automate the field data collection for characterizing motor-manual felling of willow short rotation coppices. The results were promising. By thresholding the acceleration data, the running and stopped engine states were accurately separated. Also, by combining the GPS speed with the acceleration data, followed by threshold setting and data visualization in the Geographic Information System software, detailed time categories, such as productive, working, and non-working times, could be separated. The methods described herein could be used to manage long-term field data collection, as such operations are affected by many operational factors.

Keywords: Automation; Global positioning system; Accelerometer; Thresholding; Geographic information system; Willow; Short rotation coppice; Motor-manual; Harvesting

Contact information: Department of Forest Engineering, Forest Management Planning and Terrestrial Measurements, Faculty of Silviculture and Forest Engineering, Transilvania University of Braşov, Şirul Beethoven, No. 1, 500123, Braşov, Romania; *Corresponding author: stelian.borz@unitbv.ro

INTRODUCTION

The study of time and motion is one of the most highly used research tools for evaluating the performance of various technical systems. In systems designed and implemented to procure lignocellulosic biomass for various applications, such studies are implemented to understand the behavior of the productive performance of the equipment relative to the variation in the working conditions (Visser and Spinelli 2012). The goal of these studies is to match the capabilities of the machines with their operational environments and to evaluate the environmental performance of a system or piece of equipment in terms of the energy inputs and potentially harmful outputs such as greenhouse gases emissions (Picchio et al. 2009; Balimunsi et al. 2012; Heinimann 2012; Maesano et al. 2013; Vusić et al. 2013; Ignea et al. 2016; Ackerman et al. 2017). This is done as a measure to choose the best technical option or to obtain detailed overviews of the time consumption categories required to develop work and cost rating systems (Toupin et al. 2007), which helps in the optimization of the systems in question (Ghaffaryian et al. 2010; Jourgholami et al. 2013). To this end, modeling studies are able to provide detailed elemental data (Borz et al. 2014b) and empirical relations required for optimization tasks. Similar approaches are used to perform such studies both in traditional forestry and short rotation coppice (SRC), as is shown by many recent studies.

Several methods, techniques, and procedures are currently used to conduct time studies. The traditional methods are based on manual approaches and are characterized by a series of limitations (Borz et al. 2014a; Contreras et al. 2017), including the need to use many trained people to collect field data (Borz and Ciobanu 2013; Borz et al. 2013). Such people need to focus their attention on collecting time inputs, operational variables, and production outputs. This requires a substantial measurement effort and may cause errors. Nevertheless, traditional studies are considered to be the backbone of forestry-related production studies (Heinimann 2007) and are still used in many parts of the world. In contrast, modern techniques have shown a great potential in automating parts of typical tasks in such studies. For instance, video recording approaches using devices able to capture and store digital data have enabled the automation of field data collection and can detail the real sequence of the operations being surveyed (Borz et al. 2014a; Apǎfǎian et al. 2017; Contreras et al. 2017). Video recording approaches are also used to validate the results of other approaches for studying time (McDonald et al. 2001; McDonald and Fulton 2005; McDonald et al. 2008). Nevertheless, data processing at the office, where researchers need to play the files and extract the elemental time consumption, is still labor intensive. Recent studies have shown that the amount of time spent on such tasks depends on the study design and complexity, and it can be two to six times greater than the length of the captured video files (Borz and Adam 2015; Mușat et al. 2015). In contrast, sensor-based studies are characterized by a promising potential to automate field-data collection and partly substitute the human intelligence required for separating and extracting meaningful data (Borz 2016; Cheţa and Borz 2017). Coupled with other data collection means, they can add self-data collection capabilities to a given machine or tool, including measurements of the fossil energy inputs (Talagai and Borz 2016). Usually, fossil energy inputs are used to evaluate the environmental performance of lignocellulosic biomass procurement (Picchio et al. 2009; Balimunsi et al. 2012; Heinimann 2012; Maesano et al. 2013; Vusić et al. 2013; Ignea et al. 2016; Ackerman et al. 2017), and their amount is dependent on the engine running time (Vusić et al. 2013). Therefore, by using such approaches, one could simultaneously obtain meaningful data on the operational and environmental performance of a given system. At the same time, finding ways to exclude the presence of researchers in the field has many benefits related to financial resources, data accuracy, and work safety, as well as preventing observer bias (Acuna et al. 2012).

The Global Positioning System (GPS) has the capability to document the equipment movement, and it has been used in both traditional forestry (McDonald et al. 2001; McDonald and Fulton 2005; McDonald et al. 2008; Borz et al. 2015; Strandgard and Mitchell 2015) and SRC (Eisenbies et al. 2014; Bush et al. 2015) to perform time and motion studies. While there are many solutions and devices on the market, research has shown that low-cost, consumer-grade GPS receivers are able to procure very accurate data, even when the motion speed is low (Keskin and Say 2006). Study approaches that couple acceleration sensors with a GPS system have been successfully used to document other features that could not be collected using GPS alone, for both manual (McDonald et al. 2008) and fully mechanized equipment (Strandgard and Mitchell 2015).

Because of specific operational management, the establishment of SRC plantations has more in common with agriculture than forestry (van der Meijden and Gigler 1995; Tubby and Armstrong 2002). Harvesting operations are done by implementing either a cut-and-chip or cut-and-store system (Vanbeveren et al. 2015), with the latter being performed using a chainsaw (Burger 2010; Schweier and Becker 2012) or brush cutter (Talagai et al. 2017). In general, motor-manual operations done by brush cutters have been shown to be characterized by an intensive amount of physical work (Toupin et al. 2007) and increased delivery costs (Vanbeveren et al. 2015). Nevertheless, motor-manual equipment is better fitted and is still widely used in small-scale willow SRC cut-back and harvesting operations (Talagai et al. 2017) or in cases where fully mechanized equipment is not available (Vanbeveren et al. 2015). Such contexts are typical of traditional forestry in Eastern European countries, including Romania (Rauch et al. 2015; Moskalik et al. 2017), and probably also willow SRC because such practices are relatively new in the area (Scriba et al. 2014).

When using a brush cutter to perform felling operations in a willow SRC, the traditional time and motion studying techniques are labor intensive because they require at least one field researcher (Talagai et al. 2017) to monitor the operations and capture time consumption data, which is something that can only be achieved by following the harvesting crew during the operation; this requirement may further expose the researcher to safety hazards. Additionally, the use of the snapback chronometry method (Björheden et al. 1995), which can save time during data processing tasks, is technically limited because of the prospective appearance of very short quick-changing events. Moreover, operational variables need to be collected, such as those that characterize the row lengths and biomass production (Talagai et al. 2017). These tasks require the presence of additional research personnel. At the same time, various events can characterize a time study and influence time consumption during motor-manual harvesting operations in willow SRCs. Moving with the engine running while felling, moving with the engine running without felling, headland turns, delays (at the headlands or within the land), and time spent to replace the felling discs or to refuel the tool are typical examples of such events. In particular, a separation of the running and stopped engine states is important from a time consumption standpoint because such states are likely to characterize the operation and nonoperation events. Also, the operation time is related to the fuel and energy intake.

The goal of this study was to test the capability of an acceleration sensor coupled with a low-cost GPS receiver to automatically collect meaningful time and motion data during motor-manual willow SRC felling operations using a brush cutter. In particular, the authors wanted to test the capability of the studied system to distinguish between the running and stopped engine states, and also test its capability to distinguish between specific work elements.

EXPERIMENTAL

Motor-manual felling operations using brush cutters were done on the 27th of February, 2017 in a willow SRC located near Poian, Covasna county, Romania. The owner of the SRC uses this plantation also to procure regeneration material, which means that each year he harvests smaller plots to manufacture the cuttings needed for the establishment of other SRCs. In such cases, the typical procurement operations are motor-manual felling, followed by manual bunching and hauling to a storage facility where the cuttings are manufactured. Operations were performed on a 0.5-ha plot (plantation scheme of 0.75 m between rows and 1.50 m between twin rows) located at 46° 04’ 21” N – 26° 10’ 55”, 580 m above sea level on flat land (Fig. 1), using a Husqvarna 545 RX brush cutter (Husqvarna AB, Stockholm, Sweden) equipped with a steel saw blade (Fig. 2) with a crew consisting of two men. The operations were monitored for a full working day during which the crew managed to operate all of the area designated as the study plot.

Fig. 1. Study location

Fig. 2. Placement of the dataloggers: A: general setup and B: acceleration sensor placement

An Extech® VB 300 vibration and motion detection datalogger (Extech Instruments, FLIR Commercial Systems Inc., Nashua, NH, USA) was placed on the engine of the brush cutter in a location that avoided obstructing the usual way of working. Such dataloggers are able to collect accurate (± 0.5 g) three-axis acceleration data in the range of plus or minus 18 g at a sampling rate that ranges between 500 milliseconds for online sampling and 24 h for outline sampling that is stored on a 4-Mb internal memory. A key characteristic of the datalogger used is its reduced size (95 mm × 28 mm × 21 mm) and weight (20 g), which eliminates the carrying burden and makes it feasible for motor-manual studies. Additionally, the setup and data downloading tasks were supported by dedicated software that enables the adjustment of the sampling rate, as well as data visualization and exporting to MS Excel (Microsoft, Redmond, USA). In this study, the highest available offline sampling rate (1 s) was used. A Garmin 62 STC GPS unit (Garmin Ltd., Olathe, USA) was used to document the movement data. For the sake of data processing efforts and because of the slow movement characteristics of similar operations (Talagai et al. 2017), this study used a sampling rate of 5 s for the positioning data.

Both the GPS and acceleration sensor are capable of detailed data documentation, including the use of date and time labels paired with the sampled features. This capability was used to pair the clocks of the devices prior to their setup with a portable computer. The GPS and acceleration sensor were started manually at the same time, and then placed on the brush cutter engine and strap, respectively. After placement of all of the devices, the workers were instructed to perform their tasks as usual, and the operations were videotaped as they progressed. At the end of the day, the devices were taken off of the equipment and the datalogging was switched off simultaneously on both devices.

The positioning data was downloaded to a computer as a .GPX file using regular data transfer procedures. It was then uploaded into the Base Camp software (version 4.6.2., Garmin Ltd.) for further analyzation. Processing tasks consisted of extracting the time labels, as well as the time (T, 5 s) and speed (S, km/h), for each of the collected locations as text strings. These were transferred to an MS Excel spreadsheet, where the time and speed were converted into numbers using simple MS Excel functions. The data collected by the accelerometer was downloaded using its dedicated software. Then it was exported as an MS Excel .CSV file containing the time labels and values of the acceleration (A, g) on three reference axes, as well as the vector sum of the three. Because of the different sampling rates that were set for the devices, a simple procedure was used to extract every fifth value from the acceleration data. The time labels were then used to pair the processed data from both devices.

Part of the data analysis such as characterizing the engine state and extracting the time consumption specific to distinguishable events was done in MS Excel. There are many systems that can be used to describe, characterize and categorize the inputs and outputs when studying the performance of agriculture- and forestry-related procurement systems (ASAE 2011; Lu and Ackerman 2012). This study used the IUFRO classification system (Björheden et al. 1995). In particular, the non-work and work time categories were the subject of this study, with the aim of separating the productive time from the non-work time. In the IUFRO classification system, productive time consists of the main and complementary work time categories. The main work time was assimilated in this study as the effective cutting time (CT, s), while the complementary work time (AT, s) was associated with other categories, such as walking with the brush cutter running without cutting within the plot and walking with the brush cutter stopped within the plot. All of the other categories were treated as non-work time (ST, s). The sum of all of the time categories was considered to be the total study time (TT, s). Because of the sampling strategy, this study assumed a time consumption data accuracy of ±5 s.

The first step was that of comparing the original acceleration data pool with the refined one to check for eventual data loss caused by resampling. The checking procedure consisted of extracting the engine running time data from both sets based on the acceleration behavior, followed by a percent comparison of the two. Then, to extract the time consumption data, a threshold setting procedure was used. First, the acceleration (A) data as vector sums was plotted against the speed (S) data extracted from the GPS files. Movement detection was based on a threshold set at a S greater than 0.5 km/h. The assumed threshold was documented based on the speeds reported in the previous study by Talagai et al. (2017), as well as on the figures reported by other studies related to GPS movement and speed change detection (Keskin and Say 2006; Eisenbies et al. 2014; Bush et al. 2015). Two thresholds were set for the A data. The first one was set at 1.5 g to characterize the engine non-working state. The device used in this study captures accelerations close to 1 g when detecting no vibrations. The second threshold was set to 4 g to distinguish between the engine running (TR, s) and stopped (TS, s) times in which other events occurred that caused accelerations in the range 1.5 g to 4 g, such as replacing the steel blade. The thresholds set for the GPS and accelerometer data were further used to characterize specific events, including engine stopped and no movement (A ≤ 1.5 g, S ≤ 0.5 km/h) corresponding to ST, engine stopped and movement within the plot (A ≤ 1.5 g, S > 0.5 km/h) corresponding to AT, engine stopped and movement outside the plot (A ≤ 1.5 g, S > 0.5 km/h) corresponding to ST, and cutting events (A ≥ 4 g, S > 0.5 km/h) corresponding to CT. To properly distinguish the events located outside the plot, this study also used the open-source software QGis (https://qgis.org/en/site/) to map the data. To this end, a .SHP layer was designed to store the original GPS data. It was then used to import and pair the data processed in MS Excel. The first analysis aimed to separate and map the two engine states (running and stopped) and it was performed using only the threshold set for acceleration. The second analysis aimed to distinguish between the time consumption categories and it supposed the use of a simple coding procedure in MS Excel V.B.A. to code the events as numbers and to extract the time consumption of each event. The obtained codes were imported to the .SHP file and plotted as a second map in QGis, while the time consumption data was used for further performance assessment. To evaluate the productivity and efficiency of the operations, estimates of the time consumption data (that were produced using the above described procedures) and characteristics of the plot (total row length and plot area) were used. The productivity was estimated as the ratio of operated area to time inputs, while the efficiency was estimated as the inverse ratio.

RESULTS AND DISCUSSION

This study covered more than 5400 GPS positions and almost 26700 acceleration samples. Figure 3 shows the differences between the two sampling approaches in terms of shares in the total study time (TT). Resampling the data led to similar results in terms of recognition accuracy for the two engine states (TS and TR), as the differences between the two were less than 0.01%.

Most of the time (> 75%), the engine was recognized to be in the running state, which could have indicated a high proportion of productive time (Figs. 4 and 5). Such a good recognition of the running time may help in designing further research on relating the fuel intake to the engine running time.

A detailed overview of the recognition accuracy of the original and resampled data is given in Fig. 4, which shows the engine running (TR) and stopped (TS) states based on the used threshold (A = 4 g). Preparing the brush cutter, including mounting and dismantling the sawing blades, were events that were characterized by very short durations where the acceleration exceeded 1.5 g, but it was less compared to that of the cutting activity. Also, the placement and takedown of the device on the brush cutter was characterized by an acceleration that exceeded 4 g without reaching the level specific to cutting activities.

Fig. 3. Shares of the TS and TR states in the TT depending on the sample rate