:: Industrial Engineering & Management Systems ::

• Editorial Board +
• For Contributors +
• Journal Search +
Journal Search Engine
ISSN : 1598-7248 (Print)
ISSN : 2234-6473 (Online)
Industrial Engineering & Management Systems Vol.17 No.2 pp.177-183
DOI : https://doi.org/10.7232/iems.2018.17.2.177

# Feasibility Research on Development of Product-Independent Assembly Complexity Model: A Case Study of a Refrigerator

Bongjun Ji, Hyunseop Park, Hyunbo Cho*, Kiwook Jung
Department of Industrial and Management Engineering Pohang University of Science and Technology, Pohang, Gyeongbuk, Republic of Korea
Production-based Technology Department, LG Electronics, Pyeongtaek, Gyeonggi, Republic of Korea
Corresponding author : hcho@postech.ac.kr
January 23, 2017 October 9, 2017 April 1, 2018

## ABSTRACT

The earlier studies about product assembly-complexity are validated in the specific product category. Therefore, product assembly-complexity model can be developed for specific product category and applied to validated product category. One of the earlier research on product assembly complexity model, which is developed by ElMaraghy, is validated only in the automobile engine assembly process. This paper applies the ACM to the process of assembling a refrigerator. For this purpose, the ACM is updated to represent the characteristics of this process. The assembly complexity calculated for a refrigerator is significantly correlated with assembly time. This result implies that the ACM can be made generic by updating or modifying it to represent the characteristic of a specific product-assembly process, and can be used for a wide range of product categories.

## 1. INTRODUCTION

Assembly complexity can be defined as the degree of difficulty involved in combining parts or sub-assemblies into a completed product. An assembly is an organized group of parts, and is an intermediate product of the process of assembling larger objects. The assembly can be classified into those that: 1) are composed of parts; 2) are sub-assemblies; 3) are assembled from sub-assemblies. In an early design stage, knowledge of assembly complexity is useful to the manufacturer. Assembly complexity has a strong relationship with time of execution (Fitts, 1954). Using this characteristic of assembly complexity, the manufacturer can roughly predict labor cost, and thereby improve production planning.

The number of defects decreases when manual assembly efficiency increases (Brannan, 1991). Hence, the defect rate can be reduced by decreasing assembly complexity and selecting design alternatives that have the lowest assembly complexity. Therefore, quantifying assembly complexity reasonably and objectively has been an important goal for manufacturers and researchers. This paper focuses on validation and extension of an assemblycomplexity model (ACM) of Samy and ElMaraghy (2010). when applied to a new product category. The ACM considers both handling and insertion attributes. These attributes are widely used in industry to evaluate assembly complexity, so the ACM is easily adopted in industry. Therefore, we selected the ACM for this feasibility study. ElMaraghy developed the ACM for analysis of assembly of an automobile engine, and validated the model by comparing calculated assembly complex index with assembly time obtained using Design for Assembly (DFA) analysis. However, the applicability of this model to other product categories has not been demonstrated. If this assembly complexity model can be applied without limitations to product categories, it will be a useful tool for product design and managing manufacturing process in the overall manufacturing industry. In ElMaraghy’s research, the time used in the validation was not real assembly time, but the time from DFA analysis; furthermore, the ACM was based on the DFA analysis. Therefore, the correlation of assembly complexity to assembly time is trivial, because they are both obtained from DFA analysis. In this paper, we used the ACM, but validated it by considering assembly time measured in a real assembly process.

## 2. RELATED WORK

The first representation of assembly complexity was the Predetermined Motion Time System (PMTS) in 1924 (Niebel, 1962; Hinckley, 1994; Meyer, 2008). PMTS breaks down human motion into a predetermined set of basic motions such as reach, move, grasp, position, release; the features of motion are aggregated to calculate the required motion time. Methods Time Measurement (MTM) was developed by Harold B. Maynard in 1946 (Razmi and Shakhs-Niyaee, 2008). MTM uses predetermined times for motions such as reach, move, grasp, position, disengage, and release. To predict assembly time, MTM considers both motion and the distance of motion. In 1965, MTM was updated to MTM-2 which is much simpler and less time-consuming than MTM. In 1970, MTM-3 was developed; it is simpler and less timeconsuming than MTM-2 (Salvendy, 2001).

The Maynard Operation Sequence Technique (MOST) (Zandin, 2003) focuses on the movement of the object. MOST breaks work down into general move, controlled move, and tool use. Depending on what it focuses on, MOST also has many variations such as BasicMOST (General operations), MiniMOST (Repetitive Operations), MaxiMOST (Non-Repetitive Operations) and AdminMOST (Clerical Operations).

Modular Arrangement of Predetermined Time Standards (MODAPTS) which is proposed by Stewart (2002) uses 1 MOD (= 0.129 s) as a unit measurement. The method also breaks down the motion into predetermined motions, which are coded using a letter and an integer.

Alkan et al. (2016) proposed a method to assess the process complexity of manual assembly operations. Representation of manual operations based on PMTS is used.

Although PMTS is a useful tool for decision support, the output requires much time to analyze, and can only be operated by trained experts. Those limitations led researchers to change the way in which assembly complexity is measured.

Mattson et al. (2011) developed Comple Xity Index (CXI) aims at measuring the production complexity of workstations at the operator level. Basic Assembly Complexity (CXB) assess the basic design of products, components and assembly concept (Falck et al., 2017). CXB method are developed by Falck et al. (2014) and it proposed 16 basic assembly complexity criteria.

DFA is another widely used concept in assembly complexity assessment. It considers part characteristics more than process characteristics, and therefore assesses assembly complexity easily. This method is fast and does not require well-trained expertise. DFA does not require demonstration of the assembly process, so the cost of assembling newly-designed products can be easily estimated.

The Hitachi corporation was the first developer of DFA. The DFA method that it developed is called the Assembly Evaluation Method (AEM). It estimates design coefficient (fault occurrence level) by multiplying a process factor and a parts factor. Process factors are characteristics related to assembly operation. Parts factors are characteristics of parts. Each factor has a numerical value that is based on its failure rate.

The Boothroyd and Dewhurst (1987) method is based on assembly time. Boothroyd used empirical study to identify parts characteristics that affect assembly time, and by how much. The main idea of the method is not much different from “assembly complexity,” although the description does not use the term. Therefore, Samy and ElMaraghy (2010) developed ACM (Table 1) based on the DFA. The ACM assess the assembly complexity of individual parts using an index for measuring the complexity. Index reflects assembly attributes as well as their number and variety. However, it is validated only in the automobile engine assembly process. This paper applies the ACM to the process of assembling a refrigerator for feasibility research on development of product-independent assembly complexity model.

## 3. EXPERIMENT

The objective of this experiment is to test whether the ACM can be applied to a refrigerator-assembly process. ElMaraghy classified the assembly process into two types: handling and insertion. During each process, he used DFA-based analysis to select physical and geometrical attributes that affect the difficulty of assembling the part or sub-assembly. Therefore, the assembly complexity model is similar to DFA rather than to PMTS.

Calculation of assembly difficulty factor was based on the physical or geometrical attribute. The selected attributes of each process were classified into detail levels. Each level has an average difficulty factor, which quantifies the difficulty of assembly when the process has that characteristic. Preparation process time and finishing process time (after assembly time) are most relevant to process layout and type of part, so in this experiment, assembly complexity is calculated using a unit process. Observed time measured as assembly time can vary according to its definition. In this experiment, the time from the moment of grasping the part/sub-assembly to the moment of release is defined as the assembly time (Figure 1). The scope of assembly time in this paper is the time excluded preparation and finishing process time.

The hypothesis of this experiment is that observed assembly time is correlated with calculated assembly complexity index, i.e., that the time to assemble an object increases its assembly complexity index. If the hypothesis is not rejected, then ACM can be applied to a refrigerator.

### 3.1. Data Collection and Model Update

Data were collected by analyzing (Figure 2) a video recording of an assembly process. To collect the data, 712 assembly processes involved in assembling a refrig-erator were analyzed. The values of 11 of 15 attributes were estimated by analyzing the assembly video; the values of the other 4 attributes and assembly process profile (which part is assembled) were estimated using bill of material (BOM) and a computer-aided design (CAD) file. Also, assembly time of each process was measured for validation of the ACM. The assumptions used when measuring assembly time were (Orlandi and Barnes, 1967; Helu et al., 2015; Sable, 2012):

• 1) The operator who assembles part/sub-assembly is well trained, and qualified

• 2) The operator assembles the part/sub-assembly at normal pace

• 3) The operator does a specific task

These three conditions are widely-used basic condi-tions for time study.

Because of the difference between automobile as-sembly and refrigerator assembly, the existing ACM lack in some aspect. First, assembly process in refrigera-tor assembly is not covered in automobile assembly based ACM. For example, taping is one of the most common fastening processes in refrigerator assembly process, but it is not used when assembling automobile engines. Hence, the ACM should be modified in parallel with estimating assembly complexity attributes. The difficulty factor of new level in the attributes reflect the amount of time spent and fatigue of worker. The quan-tification is based on comparison to other difficulty factor of level in same attributes. Second, the part as-sembled in refrigerator are smaller than the parts as-sembled in automobile, and it is more likely to be not fixed on product. Hence, whether the parts are placed in a stable manner or not is important in the assembly process in refrigerator. When the contact between the floor and part is surface, it is more stable than contact with line shape. These aspects should be reflected to ACM.

After collection, the raw data were refined. Alt-hough the values of all estimated attributes are the same, the measured time is not always the same. Root cause analysis was conducted on values that deviate from the mean by > 2 standard deviations; if the root cause does not meet the assumption, the values are deleted.

Difficulty factor of modified level is also calculat-ed in this step. The difficulty factor simply the time normalized by its maximum value. Video, CAD, and BOM were analyzed to estimate the values of attributes based on the modified ACM.

### 3.2. Calculation of Assembly Complexity Index

Based on collected data and assembly complexity model, complexity is calculated (Samy and ElMaraghy, 2010).

Part complexity index is calculated as

where $c h = ∑ 1 J C h , f J$, which is a handling complexity factor and $c i = ∑ 1 K c i , f K$, which is an insertion complexi-ty factor.

In the above formula, J and K refer to the number of handling attributes and insertion attributes of each part respectively. Table 2 shows the examples of collected data and its complexity factor.

The simplified asseembly process of door part is represented in Figure 3 which is assessed in Table 2.

### 3.3. Validation of Assembly Complexity Model

The relation between calculated assembly com-plexicomplexity index and measured assembly time was analyzed by calculating correlation. The Pearson corre-lation coefficient between assembly complexity index and measured assembly time was 0.61 (p-value < 0.001). High correlation indicates positive relationship between assembly time and assembly complexity. In other word, assembly time increased as assembly com-plexity index increased (Figure 4). The process who has highest assembly complexity require the longest assem-bly time. Because we defined assembly complexity as the degree of difficulty in assembly complexity, the as-sembly time is appropriate measure for assessing as-sembly complexity.

### 3.4. Feasibility of Product-Independent Assembly Complexity Model

The ACM that was developed based on automo-bile engine assembly was successfully applied to refrig-erator assembly. This result confirms the possibility of developing a product-independent ACM. Because some characteristics of refrigerator assembly do not occur in automobile assembly, a minor update to the ACM was necessary. However, even after this update, the same developing principle could be applied.

To consider an update, if the attribute has a categorical level, the level that is not included in the ACM can be added easily with no need to update the binary level. However, if the attribute has a continuous level, the choice of ranges into which it should be divided can be a problem. For example, in ElMaraghy’s ACM, difficulty factor increases as component size decreases, but the refrigerator assembly process in-volves some parts that are too big for one person to handle. Therefore, when the part exceeds a certain size, the difficulty factor should be increased. For this kind of update, expertise is needed. The same approaches can be applied other product categories.

## 4. CONCLUSION AND FUTURE WORK

Knowing assembly time is a benefit to the manufacturer because the knowledge can be used a as decision-support tool during production planning, and for managing labor cost. Minimizing defect rate is an-other topic of interest to a manufacturer. Assembly complexity, which is a degree of difficulty to assemble, has relation with both assembly time and defect rate, so many researchers have tried to quantify assembly complexity. ElMaraghy developed a successful model, but its validation was limited to the assembly of automobile engines. Furthermore, its validation was based on time from DFA analysis, not from real assem-bly time, so this paper focuses on testing whether the existing ACM can be modified to apply to a new prod-uct category, refrigerator assembly process. and to use real assembly time to test this possibility. Applying the existing model required some update, but it was appli-cable to refrigerator assembly. If the ACM can be vali-dated on additional product categories, it can be modified to be product-independent. However, assem-bly complexity tends to be concentrated in specific score range with larger difference in assembly time. Hence a further study about assembly complexity model should focus on distinguishing the different char-acteristics of assembly processes in this range.

Processes that have low assembly complexity seemed to be automated easily at low cost, because they require a low degree of precision, and simple as-sembly directions. Precision and cost will be obstacles to automation, so in this respect, ACM has potential to evolve into an automation index model.

Also at present, the manufacturing paradigm is changing. The number of products, quality requirements, and productivity should be increased to satisfy custom-er needs. For the technology aspect, the smart manufac-turing concept is developing. Hence, some factories use Model Predicted Control (MPC) to optimize their manu-facturing environment. However, MPC cannot be easily applied to labor-intensive manufacturing, because the human factor is difficult to quantify and control. Hence controlling the other controllable factors like assembly complexity can be an alternative.

## ACKNOWLEDGEMENTS

This research was partly supported by the Smart Fac-tory R&D program of KEIT [10054495 Development of data collection/processing systems capable of adapt-ing manufacturing environment and building a site for demonstra-tions].

## Figure

Components of whole assembly time and some examples of each process.

Data collection framework.

Simplified refrigerator door assembly.

Assembly timie versus assembly complexity index.

## Table

Assembly complexity model (Samy and ElMaraghy, 2010)

Examples of collected data and its complexity factor

## REFERENCES

1. B. Alkan , D. Vera , Mr. Ahmad , B. Ahmad , R. Harrison (2016) A model for complexity assessment in manual assembly operation through predetermined motion time systems., Procedia CIRP, Vol.44 ; pp.429-434
2. G. Boothroyd , P. Dewhurst (1987) Product design for assembly., Boothroyd Dewhurst Incorporated,
3. B. Brannan (1991) Six sigma quality and DFA-DFMA case study/motorola Inc., Boothroyd & DewhurstDFM Insight, Vol.2 ; pp.1-3
4. A. Falck , M. Tarrar , S. Mattsson , L. Andersson , M. Rosenqvist , R. Söderberg (2017) Assessment of manual assembly complexity: A theoretical andempirical comparison of two methods., Int. J. Prod. Res., Vol.55 (24) ; pp.1-14
5. A. Falck , R. A-rtengren , M. Rosenqvist (2014) Assembly failures and action cost in relation to complexity level and assembly ergonomics in Manual assembly (part 2)., Int. J. Ind. Ergon., Vol.44 (3) ; pp.455-459
6. P.M. Fitts (1954) The information capacity of human motor system in controlling the amplitude of movement., J. Exp. Psychol., Vol.47 (6) ; pp.381-391
7. M. Helu , K. Morris , K. Jung , K. Lyons , S. Leong (2015) Identifying performance assurance challenges for smart manufacturing., Manuf. Lett., Vol.6 ; pp.1-4
8. C.M. Hinckley (1994) A global conformance quality model. A new strategic tool for minimizing defects caused by variation, error, and complexity, Ph.D.dissertation, Sandia National Labs, Livermore, CA.,
9. S. Mattson , P. Gullander , A. Davidsson (2011) Method for measuring production complexity, Proceedings of the 28th International ManufacturingConference, Dublin City University,
10. Meyer (2008) Time Study Method Implementation in Manufacturing Industry, A B.E Report, Universiti Technikal Malaysia, Melaka.,
11. B.W. Niebel (1967) Motion and time study., RD Irwin,
12. B. Orlandi , R.M. Barnes (1967) Motion and Time Study: Design and Measurement of Work., Wiley & Sons,
13. J. Razmi , M. Shakhs-Niyaee (2008) Developing a specific predetermined time study approach: An empirical study in a car industry., Prod. Plann. Contr., Vol.19 (5) ; pp.454-460
14. S.R. Sable (2013) A study of labor productivity through application of Stopwatch time study method andmost., Skoda Auto India Private Limited,
15. G. Salvendy (2001) Handbook of industrial engineering: Technology and operations management., John Wiley & Sons,
16. S.N. Samy , H. El , H. Maraghy (2010) A model for measuring products assembly complexity., Int. J. Comput. Integrated Manuf., Vol.23 (11) ; pp.1015-1027
17. J. R. Stewart (2002) Applying MODAPTS Standards, The Society for Work Science, Fall, 1,
18. K.B. Zandin (2003) MOST: Work Measurement Systems (3rd ed.), Boca Raton, Florida.,
 오늘하루 팝업창 안보기 닫기