Keywords

1 Introduction

In recent years, the depletion of natural resources has been severe globally. One of the solutions to this problem is to reuse and recycle materials from End-of-Life (EOL) products and reduce waste [1]. In the industrial field, disassembly work is needed to take back parts/materials from EOL assembly products. Furthermore, in the manufacturing industry, the labor shortage due to the falling birthrate and aging population has made it difficult to train human resources and achieve skill transfer [2]. In order to solve such problems, the manufacturing industry has been promoting digitization using sensors and Artificial Intelligence (AI) analysis [3, 4]. Digitization has a profound impact on daily life, changing people’s lives for the better in all aspects. Such a digitization method is called “Digital Transformation (DX)” [5]. In addition, motion capture has been introduced at the work site for more effective instruction and skill transfer [6]. Motion-capture [7] technology can accurately measure human body movements, and can acquire time and three-dimensional time-series coordinate data simultaneously.

In previous studies, human positioning measurements have been performed at work sites using acceleration and magnetometer sensors [8]. Liu et al. [9] conducted research using video motion capture to realize a manufacturing system in which robots and human coexist. However, optical motion capture, which could accurately measure position coordinates was not used. On the other hand, Kawane et al. [10] measured and analyzed the assembly work using optical motion capture and machine learning, although they did not treat the disassembly works. Moreover, Wilhelm et al. [11] proposed an ergonomic approach to optimize an entire manual assembly process chain in a manual assembly line using ErgoSentinel Software and a Microsoft Kinect depth camera. However, the measurement of disassembly works and that with optical motion capture, which can obtain high-precision absolute position coordinates, are not conducted.

This study measures and analyzes the motion data for disassembly work obtained via optical motion capture, and the following Research Questions (RQs) are developed.

  • RQ1: How do the worker’s movements change with each disassembly work?

  • RQ2: How does the worker proficiency affect the disassembly work?

2 Method

This study designs an experiment, where the disassembly process involves loosening nuts. The nut-loosening work is conducted with the instruments on the desk with standing, and loosens the nut with their dominant hand while holding the instrument with the opposite hand. The nuts are of the M16 type [12], i.e., non-loosening nuts.

The measurement is performed using an MAC3D Kestrel-300 optical motion capture system from NAC Image Technology, Inc. [13] at a frame rate of 60 fps. The accuracy of the MAC3D system is less than 1mm average error [13], and the least significance of the raw data is 10 nm displayed in the device. The amount of data for 3D coordinates obtained via optical motion capture is large, and an operation for approximately 15 s consists of approximately 36,000 data (approximately 1,000 frames × 3 dimensions × 12 markers) as in the case of this study. The data measurement procedure is as follows.

  1. 1.

    The worker wears clothing with 10 reflective markers (Fig. 1).

  2. 2.

    A wrench (Fig. 1) with reflective markers o the upper jaw and grip is used.

  3. 3.

    The nut is loosened for two turns by using the wrench according to a line marked in advance.

  4. 4.

    The third work of 10 times and a 5-min break is conducted three times.

Figure 1 shows the positions of the reflective markers attached to the operator and the wrench based on Helen Hayes Marker Set [14]. For the worker, Three-Dimensional (3D) position coordinates are obtained for the top of the head, back of the head, front of the head, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, and back. For the wrench, the 3D position coordinates of the upper jaw and grip are acquired.

To analyze the motion changes and the worker proficiency in the disassembly work by obtaining time-series coordinate data via motion capture, the average, standard deviation and the movements of each disassembly work are compared.

Fig. 1.
figure 1

Attachment points of reflective markers for using motion capture.

3 Results

For considering a difference by the worker’s proficiency, the working time and the values obtained via motion capture in each working session are compared in this section.

3.1 Result of Disassembly Time

Figure 2 shows the observed working time of disassembly work from the 1st to the 30th operation for loosening the nut. The horizontal axis indicates the work number, and the vertical axis shows the working time at each work. The yellow lines correspond to intervals of respective 10 operations, and the red line indicates the mean working time. It is found that all the working times from the 21st to the 30th works were shorter than the average working time.

Table 1 presents the mean and standard deviation of working time for each of the 10 work sessions. The average working time for session 3, consisting of the 21st to the 30th operation, was reduced by 25.3% and 29.4% compared with one in the other two sessions, respectively. Thus, this result implies learning effect, where the workers became accustomed to the operation through repetition. The standard deviations for the session 2 and session 3 are greater than ones for session 1. It is considered that worker fatigue and the search for tricks to perform the work efficiently have caused the variation in the working time. In addition, the learning effect observed and expected level since the value of session 3 became smaller than that of session 2.

From now, to identify the differences in the movements that affected the reduction in the working time, the transitions of the 3D coordinates for each marker in the first two operations and the last two ones are observed and compared.

Fig. 2.
figure 2

Observed working time of disassembly work in each session.

Table 1. Average and standard deviation of working time in each session and total

3.2 Results of Movements Measured

In the previous section, it is found that there is a possibility that the worker’s physical movements are also different because the average working time was different in each session. In this section, the worker’s movements are analyzed by using optical motion capture. Specifically, to investigate the relationship between the working time and the worker’s movements, the movements of the first two operations at the 1st and 2nd and the last two operations at the 29th and 30th are compared.

Figure 3 presents the position coordinates of the reflective marker affixed to the top of the worker’s head. The horizontal axis means Y or Z coordinate value, and the vertical axis shows the number of frames at each work. As shown in Fig. 3 (1), there is a large difference in the Y-coordinate between the first two operations and the last two ones. For instance, the variation of the Y-coordinate for the last two operations was approximately by 65.3 [mm] larger on average than that for the first two operations. This indicates that the head was moved in front of the body with moving their head to forward at the same time when the worker loosened the nut in the last two operations. For the Z-coordinate, the first two movements exhibited only small fluctuations of 88.2 [mm] on average, whereas the last two movements indicated large fluctuations of 133.4 [mm] on average. This means that the worker moved his head forward, and that his center of gravity was in front of his body when he lowered his head while loosening the nut. One of the reasons is that these movements make it easier to apply force to the arms.

Fig. 3.
figure 3

Time-series data values of top of the head: (1) Y-coordinate and (2) Z-coordinate

Figure 4 shows the movements of the right shoulder, right elbow, and right wrist in the Y-Z (depth-height) plane. The horizontal axis indicates Y coordinate value, and the vertical axis shows Z coordinate value at each work. The range of motion was relatively wider for the 30th work at all three locations. The 30th movement was by 29.8 [mm] greater downward and by 116.2 [mm] greater forward for the shoulder movement. This is because more weight could be applied to the front. In contrast, the elbows and wrists moved more upward by 105.0 [mm] on average. This means that the worker turned the nut using not only the force to push it forward but also another force to move it upward. Thus, in this nut-loosening operation, putting the body weight forward and using the force to raise the elbow and wrist improved the efficiency.

Fig. 4.
figure 4

The time series data of the right shoulder, right elbow, and right wrist in the Y-Z (depth-height) plane.

3.3 Discussion

Based on the results in Sects. 3.1 and 3.2, the RQs developed in Sect. 1 are answered as follows:

  • Answer of RQ1: How does the worker’s movements change with each disassembly work?

It was found that the worker moved his head and right arm more forward by approximately 65.3 [mm] and 105.5 [mm] on average at the 30th work compared with the 1st work. Therefore, it is considered that the worker became accustomed to the operation through repetition.

  • Answer of RQ2: How does the worker’s proficiency affect the disassembly work?

The upper half of the body was moved more forward, the head and shoulder were moved more downward, and the wrist and elbow were moved more upward. These movements are considered to make the nut loosening process more efficient. In addition, the worker’s proficiency through repetition of the same work was found to reduce the working time by approximately 27% in the nut-loosening work.

The results indicate that variation in working motion and time in experimented disassembly works, can be measured and identified by using optical motion capture. Since the identified disassembly motion and time enable us to improve the work, the proposed measurement of disassembly work using optical motion capture can contribute efficient reuse/recycling and automation of disassembly works for sustainability.

The limitation of this study is that it is impractical to draw and analyze the transitions of position coordinates graphically since the data obtained via optical motion capture is huge. Thus, it is expected that machine learning, which is useful for processing large amounts of data, will be used to analyze the data. The usage of machine learning to analyze high-precision position-coordinate data obtained from optical motion capture is expected to support the transfer of operator skills from an unprecedented perspective.

4 Summary and Future Studies

This study measured and analyzed the motion data for disassembly work obtained via optical motion capture. As an example of disassembly work, the work of loosening a nut was measured and analyzed. Among a total of 30 works, the working time was relatively short for the last 10 works. Therefore, to investigate the relationship between the working time and the worker’s movements, the movements in the 3-D position coordinates obtained via motion capture for the first two works and the last two works were compared.

Future works should verify the findings from these results in every case, and consider the analysis of the motion data obtained via optical motion capture by machine learning to quantitatively extract tacit knowledge in the work [10].