The current issue and full text archive of this journal is available at www.emeraldinsight.com/0263-5577.htm IMDS 108,6 Effects of product design on assembly lines performances A concurrent engineering approach 726 Received 26 November 2007 Revised 26 February 2008 Accepted 11 March 2008 Antonio C. Caputo and Pacifico M. Pelagagge Department of Mechanical, Energy and Management Engineering, University of L’Aquila, L’Aquila, Italy Abstract Purpose – The paper’s aim is to assess the impact of product related features on the performances of assembly line manufacturing systems, also providing a specific Design for Manufacturing and Assembly rating index to assess the goodness of a product design solution with respect to assembly line performances. Design/methodology/approach – A computer simulation-based parametric analysis was carried out to assess the impact of four major product-related parameters. 216 different assembly line balance problem instances were evaluated. Findings allowed to develop a DFMA rating index specific for assembly line manufacturing as well as design guidelines. Findings – Assembly sequence degrees of freedom and the ratio of the average task duration to the maximum duration are the most influencing parameters. While the former should be maximized, only a moderate task duration variability was found beneficial. The influence of other factors resulted less marked and changing on a case-specific basis. Research limitations/implications – Complex interactions between product design features and line performances prevent generalization. The performed numerical experimentation, although extensive, remains somewhat limited respect all possible practical situations. The proposed rating index should be utilized while maintaining an overall perspective about the mutual influence of all parameters. Some suggested guidelines imply a trade off with traditional DFMA guidelines. Practical implications – Product designers are given useful insights, tools and guidelines to develop better producible products. With the proposed ranking index a designer can easily rate his choices when selecting assembly tasks and sequences, as well as rank alternative product designs solutions. Originality/value – The paper presents an original discussion about the impact of product design choices on assembly line performances. The developed DFMA rating index and guidelines are new. Keywords Design, Assembly, Assembly lines, Product design Paper type Research paper Industrial Management & Data Systems Vol. 108 No. 6, 2008 pp. 726-749 q Emerald Group Publishing Limited 0263-5577 DOI 10.1108/02635570810883987 1. Introduction Product innovation and the ability of quickly developing products and bringing them to the market are key factors in maintaining globally competitive a manufacturing company (Drejer, 2008; Hanninen and Kauranen, 2007; Khan et al., 2007; Lee, 2007; Smith and Reinertsen, 1991). During new products development, the simultaneous involvement of different functions and overlapping of their activities, along with better communication between departments, has proven to reduce development time, reduce cost and increase quality giving rise to the concurrent engineering concept (Hundal, 1997). In fact, traditionally, design and manufacturing activities have taken place sequentially rather than simultaneously leading to inefficient and time consuming iterations between design and manufacturing stages (Shukor and Axinte, 2007). Concurrent engineering philosophy (Nevins and Whitney, 1989; Parsaei and Sullivan, 1993; Smith, 1997) advocates, in fact, to carry out simultaneously the product and process design with the aim of minimizing the product life cycle cost and the time to market while providing high added value products for the consumer (Prasad, 1996; Seetharaman et al., 2007). In this framework, Design for Manufacturing and Assembly (DFMA) techniques are widely employed (Boothroyd et al., 1994; Gupta et al., 1997; Bralla, 1999). However, the traditional DFMA approach essentially focuses on obtaining a product with a high level of manufacturability. DFMA, in fact, attempts at minimizing production cost through simplification of product structure mainly resorting to a reduction of parts count, proper selection of the best combination of materials, geometry and cost-effective manufacturing methods for all parts, and simplification of manual assembly tasks. This implies that most DFMA techniques rate a product design on the basis solely of direct manufacturing and assembly costs. No further reference is given instead to the overall impact that the manufacturing tasks and product structure have on the performances of the entire production system, including planning and control issues. In fact, the review of literature in this field shows an absence of a methodology for effectively incorporating the concerns of production into the early design stages, and that most of the design for manufacturing approaches do not consider constraints related to operations of manufacturing systems, but rather deal with manufacturing processes (Govil and Magrab, 2000; Kusiak and He, 1997). This is quite a limitation as many product design variables, such as tolerance levels, assembly tasks sequence constraints, utilization of bottleneck resources, degree of variability of materials flows and process times, etc. significantly affect design, operation and management of the manufacturing system, with direct consequences on WIP levels, lead times and machines utilization (Bramall et al., 2003; Caputo and Pelagagge, 2006; Corti and Portioli-Staudacher, 2004; Govil and Magrab, 2000; Kusiak and He, 1997; Soundar and Bao, 1994). All these factors bear a cost as well as an adverse effect on system performances, which traditional DFMA techniques fail to account. In order to contribute to a solution of this problem, in a previous work Caputo and Pelagagge (2006) proposed an innovative design for production (DFP) methodology relying on a rating index which allows to carry out a more exhaustive comparative ranking of design alternatives based on manufacturing system performances as affected by product features. The index can also be utilized to supplement and extend traditional DFMA ranking techniques by looking at parameters previously neglected. In this work the attention is focused instead on assembly lines production systems in order to assess the impact of product related features on balancing efficiency and consequently provide a rating index useful to evaluate the goodness of a product design solution with respect to assembly line performances. As a result some guidelines for designers are also given. The paper is organized as follows. At first, a general discussion about assembly process complexity as affected by design choices and the impact on line performance measures is carried out supported by the relevant literature. Then four influencing parameters are selected, expressing the task duration variability, their relationship with the cycle time, the number of operations and the strength of constraints in the assembly sequence. A parametric analysis is then carried out to assess the effects of Product design on assembly lines 727 IMDS 108,6 the considered parameters on the performances of the assembly line. A thorough analysis of results follows to understand the role of each investigated parameter, to verify its degree of influence and determine the preferred value. Results of this analysis are summarized into a set of guidelines for product designers. Finally, a performance index is proposed to quantitatively assess the goodness of a product design solution with respect to the assembly line performances. 728 2. Problem statement Assembly lines are flow-line production systems consisting in a serial arrangement of N workstations through which workpieces flow at a steady pace. At each station a certain part of the total work necessary to manufacture the product is performed. The whole assembly process is made of a number of separate operations (tasks). Each task is characterized by a task duration Ti, while a set of precedence constraints, imposed by technological restrictions, usually define the ordering in which the tasks may be performed. Line throughput dictates the maximum time available at each station to perform the allotted tasks (i.e. cycle time, TC). The assembly line balancing problem involves the assignment of tasks to the workstation so that no precedence constraint is violated and the sum of operations duration at each workstation does not exceed the allowed cycle time, while optimising some performance measure so that the line operates as efficiently as possible. Most commonly the objective is maximizing the capacity utilization of the line which, in case of deterministic operation times and single-model production, may be simply expressed as minimize the number of stations for a given cycle time, minimize the cycle time for a given number of station, minimize the sum of idle times at the stations or the percentage of idle times, equalize the levels of capacity utilization at the stations and minimize flow time. However, cost and profit oriented goals are also widely adopted. Therefore, common performance measures to be optimized are the line efficiency h (percent) P Ti h¼ ð1Þ N TC which measures the capacity utilization, the balance delay time, BD, which measures the unused capacity and is equal to the sum of idle times at the stations: X BD ¼ NT C 2 Ti ð2Þ the balance delay ratio BDR (percent), being h ¼ 1 2 BDR, measuring the percent capacity inutilization: P NT C 2 T i ð3Þ BDR ¼ NT C and the balance efficiency BE (percent) measuring the equality of distribution or workload among the stations: P jS k 2 S AVG j BE ¼ 1 2 ð4Þ N S AVG where Sk is the total work time at k-th station and SAVG the average station workload. The literature concerning assembly lines balancing issues and techniques is very large and the interested reader is referred to Scholl (1999) for an in depth analysis. However, the level of balance efficiency one can expect to achieve in real life applications depends from the complexity of the actual assembly problem instance which, in turn, is largely dictated by the product design choices. In fact, the following product characteristics are likely to affect the effectiveness and efficiency of line balancing because they impact on the computational problem complexity (Scholl, 1999; Driscoll and Thilakawardana, 2001): . Number of tasks. The higher the number n of tasks to be performed the higher will be the number of workstation and the number of feasible task sequences to be explored (neglecting precedence constraints there are n! feasible task sequences). Therefore, the problem complexity is expected to grow exponentially with increasing n, but the higher number of possible sequences may increase the likelihood of finding an efficient one. . Task times. When few precedence constraints apply, a high variability of task times may enable better combination of tasks to form station loads with lower idle times respect the case of almost similar operation times. When instead the assembly sequence is strongly constrained a low variability of task times may improve the load uniformity among workstations. . Task times and cycle time. If task times are small respect cycle time there should be more station loads with low idle times because it is easier to combine small items than large ones to fit within the cycle time constraint. . Precedence constraints. In general the higher the number of precedence constraints the lower will be the number of feasible solutions, which might reduce the computational complexity, but less likely will be to find an efficient solution. Additional useful insights come from the work of Kilbridge and Wester (1961) who early recognized the relationship among the various problem parameters. According to their experimental work comparing the balance efficiency of four different products manufactured by distinct companies, it was verified that balance delay generally has a sharp increase at very small cycle times and monotonically decreases when cycle time increases, except in one case, characterized by some peculiar restrictions to the placement of fixed facilities. In that case a cycle time minimizing the balance delay was found. However, no significant impact of the actual tasks duration distribution was observed in the examined cases. The study of Kilbridge and Wester also pointed out that care should be taken in choosing a cycle time that divides evenly into the total work content time. If this relationship is not considered, and a cycle time is arbitrarily chosen, the resulting balance delay may be high, regardless of the size of the cycle time. Their evidence thus suggests that high balance delay is associated with a wide range of work element times, a high degree of inflexible line mechanization and the indiscriminate choice of cycle time. To assess such issues in a quantitative manner a number of Authors tried to present simple numerical measures of complexity for the assembly line balance problem (Talbot et al., 1986). Product design on assembly lines 729 IMDS 108,6 730 Mastor (1970) proposed the order strength OS parameter, i.e. the ratio of the number of ordering relations that exist in the assembly graph to the possible number that could exist, to measure the relative strength of precedence relations. Problems with a large OS are basically expected to be more complex than such with small OS values. However, this parameter only describes the number of precedence relations but not their structure. Dar-El (1973) introduced two distinct indices, the Flexibility ratio FR ¼ 1 2 OS, and the West ratio, WR, as the average number of tasks per station. The latter criterion, however, needs to produce a balanced line before WR can be evaluated. Problems with small FR and WR tend to be more complex. Wee and Magazine (1981) introduced the time interval TI ¼ [Tmin/TC, TMAX/TC] as the interval bounded by the ratios of minimum and maximum task duration to the cycle time, to account for the range of task times respect the cycle time. A small length of TI indicates that task times vary in a small range. Therefore, problems are expected to be increasingly complex if TI is small and is near to the right border of [0, 1] meaning that the average task duration is similar to the cycle time. However, this index neglects the distribution of actual task times. Scholl (1999) describes the time variability ratio as TV ¼ Tmax/Tmin. This has the advantage of not mentioning the cycle time which makes it an intrinsic measure of the time structure of the precedence graph rather than of single problem instances. Problem complexity is expected to grow with decreasing values of TV, which indicate that operations times vary in a small range or that the minimum task time is large. Portioli (1999) introduced the Parallelism Index: IP ¼ 1 n Li þ U i ‡ n i¼1 2 ð5Þ to measure the average number of degrees of freedom (DOF) available in the assignment of a task to a station given the precedence relationships existing among tasks. In equation (5) n is the number of operations and Ui the number of DOF at step i when the assembly sequence is built by selecting at each step the operation that yields the largest number of DOF. Li is the number of DOF at step i when the sequence is built by selecting at each step the operation that yields the smallest number of DOF. It can be expected that the higher the DOF the better because a more effective line balancing is possible. Resorting to simulation experiments Portioli shows that given a number of operations and an average task time the percent idle time reduces when both the PI and the cycle time increase because the higher DOF and smaller task times enable a better fitting assignment of tasks to stations. The same happens when the PI is fixed but average task time is reduced respect the cycle time. Johnson (1988), however, observes that the above measures only describe a single characteristic of the problem at a time and advises to choose a combination of measures, while Scholl (1999) points out that a suitable measure of complexity should incorporate information about the structure of the precedence graph together with the task times. To this end Driscoll and Thilakawardana (2001) introduce an aggregate precedence index: PI ¼ PS þ PB 2 ð6Þ where PS is the precedence strength index defined as PS ¼ (c 2 1)/(n 2 1) being c the number of precedence columns in the precedence matrix and n the overall number of tasks, while PB ¼ Cav/c is the precedence bias, being Cav the average element column position. PS represents the constraints on element selection during balancing, with extremes represented by weakly ordered diagram with no constraints and strongly ordered diagram totally sequentially constrained. PB is the precedence bias and represents the existence of available elements for assignment early in the balancing process and is a measure of the precedence diagram variability. To account for task time variability in reference to cycle time, instead, Driscoll and Thilakawardana (2001) introduce the task time index: TTI ¼ TR þ TD 2 ð7Þ where TR ¼ TAVG/TC is the ratio of average task duration to cycle time: 2sT TD ¼ 1 2 TC ð8Þ is the task time distribution index with sT being the standard deviation of task times. A high value of TR means an average task duration similar to cycle time which makes balancing more difficult, while a small value of TD means a great variability of task times with the presence of small work elements improving the ability to “pack” tasks into stations. Overall TTI and PI may be combined into a compound project index Pjt I ¼ (PI þ TTI)/2. Summing up it can be stated that intuition and empirical evidence might suggest that the performances of an assembly line production system in general are more likely to improve when: . the number n of assembly operations is large (the number of stations grows but better balancing occurs thanks to the greater choices when assigning tasks to stations and the greater number of alternative sequences to be explored); . tasks duration is highly variable, with a strong presence even of tasks of short duration (except when the assembly sequence is strongly ordered); . the average task duration is small respect cycle time; and . the assembly sequence has many DOF. This also means that an increased computational complexity, as far as the specific problem instance is concerned, often may lead to an improved balancing efficiency. However, while assembly line designers are given such input data by the design bureau or the process planners, product designers instead have the opportunity of shaping the value of such influencing parameters enabling improved performances of the manufacturing system. Therefore, DFMA techniques should be improved in order to explicitly account for assembly line designers needs. As an example in a product design the tasks could be subdivided into smaller tasks to reduce the average task time respect cycle time, more task duration variability could be designed and more flexible assembly sequences could be devised. Product design on assembly lines 731 IMDS 108,6 732 3. Analysis methodology To assess the impact of the product design features of the performances of an assembly line the following methodology was adopted. At first four distinct product design features were chosen, as representative of the complexity of assembly line balancing instance, namely: (1) the number n of assembly tasks to be performed; (2) the average number of DOF in the assignment of a task to a station given the precedence constraints in the assembly sequence; (3) the ratio of average task time to the maximum task time TAVG/TMAX (Figure 1); and (4) the ratio of the maximum task time to the cycle time TMAX/TC (Figure 1). The DOF represents the average number of choices one has in selecting the next task to be assigned to a station after a given task (i.e. a node in the precedence graph) has been assigned. It is computed as expressed in equation (5). It may vary in the interval: h Xn21 i i =n ð9Þ 1; 1 þ i¼1 T Overall the four selected variables address all of the parameters recognized as influencing in the literature on assembly line balancing (Dar-El, 1973; Driscoll and Thilakawardana, 2001; Johnson, 1988; Mastor, 1970; Portioli, 1999; Talbot et al., 1986), i.e. the number of tasks, the order strength, the task duration variability, the ratio of task times to the cycle time. Furthermore, they may be considered as equivalent to the various indices utilized by other authors to quantify such parameters. In fact parameters TAVG/TMAX and TMAX/TC may be considered to be equivalent to parameters TD and TR of Driscoll and Thilakawardana (2001) and to the time interval of Wee and Magazine (1981), while the ratio TMAX/TC has been explicitly considered by most of the cited authors. Even the task number has been generally adopted in the literature. The average number of DOF, measured through the index IP of Portioli (1999), may be considered roughly equivalent to index PS of Driscoll and Thilakawardana (2001) or to OS of Mastor (1970) and FR suggested by Dar-El (1973). However, while most authors utilized only a subset of such indices at a time, only Driscoll and Thilakawardana (2001) utilized, as happens in this work, a complete set of parameters. Furthermore, while other product features could be certainly utilized Figure 1. Sample tasks duration distribution and indication of characterizing parameters ∆T TMAX TC TAVG Tasks to characterize an assembly problem instance, i.e. tolerance levels, task complexity, buffers availability, technical constraints of work stations and so on, such specific factors are best utilized for detailed characterization of specific situations, but are not amenable to provide generalised results. Therefore, they will not be considered in the following analysis. An experimental campaign was then carried out by changing the values of each parameters, one at a time, in order to consider all combinations and evaluate the assembly line performances basing on the following measures: line efficiency h, balance efficiency BE, and the ratio of the actual to theoretic number of stations SR ¼ N/NT. As far as the number of task is concerned three levels were considered, four levels were considered for DOF, three levels were assumed for the ratio of average to maximum task time, and six levels for the ratio of the maximum task time to the cycle time. It should be noted that TAVG/TMAX and TAVG/TC cannot be changed independently because the further constraint TMAX # TC must be satisfied. Then, K 2 ¼ T AVG/T C , the condition indicating K1 ¼ T AVG /TMAX, TMAX/TC ¼ K2/K1 ¼ K3 # 1 translates in the condition K1 $ K2. The resulting experimental matrix is shown in Table I, where the adopted parameters values give rise to the following values for the TAVG/TC ratios: 0.16, 0.2, 0.24, 0.3, 0.32, 0.36, 0.4, 0.48, 0.54, 0.6, 0.64, 0.72, 0.8. The selection of levels for the examined parameters has been made in order to uniformly span across the entire variability range allowed by the parameters, but limiting their number to a manageable value in order to avoid a combinatorial explosion of the number of experiments to be carried out. The values of DOF are instead concentrated in the first half of the variation range because in that area the line balancing efficiency degrades rapidly and it is necessary to make a more detailed analysis. The levels values were chosen to be comparable with those assumed by most other authors and are expressed in relative manner for sake of generality except the number of tasks. In greater detail the maximum value of task number, n ¼ 30, compares well with that of Mastor (1970) which ranged between 20 and 40 and Portioli (1999) who assumed 20 tasks. Dar-El (1973) and Talbot et al. (1986) chose slightly larger values between 50 and 100 tasks, while only Johnson (1988) examined a very wide range, from 20 to 1000. As far TMAX/TC is concerned Talbot et al. (1986) examined values from 0.5 to 1, while Johnson (1988) from 0.1 to 0.6 (expressed as station to work element ratio). With reference to the strength of precedence relations most authors (Johnson, 1988; Mastor, 1970; Talbot et al., 1986) adopted OS values roughly in the range 0.2-0.8. Overall 216 different combinations of parameters values were examined. Actually 3 £ 4 ¼ 12 different assembly processes were compared each characterized by a different precedence graph obtained increasing in three levels the number of nodes from 10 to 30 and for each number of nodes changing the precedence constraints in N DOF TAVG/TMAX TMAX/TC 10; 20; 30 DOF min; 0.25 DOF max; 0.5 DOF max; DOF max 0.4; 0.6; 0.8 0.4; 0.5; 0.6; 0.8; 0.9; 1 Product design on assembly lines 733 Table I. Experimental matrix showing the assumed valued of examined parameters IMDS 108,6 734 order to span the entire DOF variability range. Then for each precedence graph the line balancing was carried out for all 18 possible combinations of TAVG/TMAX and TMAX/TC ratios. This two-step approach was justified considering that the number of operations and DOF define intrinsically the complexity of the instance as well as the precedence graph. Then the subsequent variation of time-related parameters explores the effect of changing the distribution of task duration and the externally imposed productivity goals (i.e. through cycle time variations). To solve each problem instance a computer model was developed performing line balancing adopting the ranked positional weight technique (Helgeson and Birnie, 1961). For each of the 216 examined problem instances, 100 replications, obtained by randomly changing task times (while respecting the specified TAVG/TMAX and TMAX/TC ratios), were utilized to compute average values of the performance measures in order to gain statistical consistency. 4. Parametric analysis results 4.1 Effect of operations number In the examined range of operations number (10-30) line efficiency h shows a steady increase as the number operation grows and in general the improvement is more marked the greater is the DOF of the precedence graph and the lower is the TAVG/TMAX ratio. This confirms previous literature findings and may be explained by the greater average number of choices of feasible tasks to saturate the workstations when n (and consequently the absolute value of the maximum DOF), increases especially when the average task duration is short respect cycle time. As an example Figures 2 and 3 show the effect of task number on line efficiency for two values of average DOF. As far as the balance efficiency BE is concerned, in average a similar improvement trend with increasing n was observed although less marked. Passing to the SR ratio instead, mixed results were obtained depending on the other parameters values, making any generalization difficult. The main finding was that SR might improve when n increases but only when DOF is large and the TAVG/TMAX ratio is small. As an example Figures 4 and 5 show the trend of SR values when the number of tasks increases, with the maximum DOF value and two values of the TAVG/TMAX ratio. 1.00 Efficiency 0.90 TMAX = 0.4 Tc TMAX = 0.5 Tc TMAX = 0.6 Tc TMAX = 0.8 Tc TMAX = 0.9 Tc TMAX = Tc 0.80 0.70 Figure 2. Effect of tasks number on line efficiency for different DOF DOF = 0.25 MAX TAVG 0.4 TMAX 0.60 10 20 N 30 1.00 Product design on assembly lines TMAX = 0.4 Tc TMAX = 0.5 Tc TMAX = 0.6 Tc Efficiency 0.90 TMAX = 0.8 Tc TMAX = 0.9 Tc 735 TMAX = Tc 0.80 0.70 DOF = MAX TAVG = 0.4 TMAX 0.60 10 20 30 N Figure 3. Effect of tasks number on line efficiency for different DOF 1.60 TMAX = 0.4 Tc 1.50 TMAX = 0.5 Tc TMAX = 0.6 Tc 1.40 TMAX = 0.8 Tc TMAX = 0.9 Tc SR 1.30 TMAX = Tc 1.20 1.10 DOF = MAX TAVG = 0.4 TMAX 1.00 0.90 10 20 30 N Computed results also show that passing from n ¼ 10-30, all the line performances increase more markedly when TMAX/TC ¼ 0.5 respect all other cases. This is a misleading circumstance caused by the value assumed by the [(TAVG £ n)/TC] ratio which defines the theoretic minimum number of stations. When the theoretic minimum number of stations is an integer any slightest balancing inefficiency causes the need for adding at least one more station with an abrupt decrease of line performances. Instead when the theoretic minimum number of stations is obtained from rounding in excess a real number it is easier to maintain the actual number of station equal to the minimum one or to have a better balancing efficiency. A synthesis of the obtained results is depicted in Table II. 4.2 Effect of degrees of freedom With reference to the influence of the number of DOF in the assembly sequence, both line efficiency h and the SR parameter steadily improve with an increase of DOF Figure 4. Effect of tasks number on SR for different DOF IMDS 108,6 1.60 TMAX = 0.4 Tc DOF = MAX TAVG = 0.8 TMAX 1.50 TMAX = 0.5 Tc TMAX = 0.6 Tc 1.40 TMAX = 0.9 Tc SR 736 TMAX = 0.8 Tc 1.30 TMAX = Tc 1.20 1.10 1.00 Figure 5. Effect of tasks number on SR for different DOF 0.90 10 20 30 N DOF Table II. Overview of performance results when operations number increases Line efficiency h Balance efficiency BE Station ratio SR Low High " " , " " " " " TAVG/TMAX Low High " " " " " " " , TMAX/TC Low High , , , , , , Notes: " ¼ improves; " " ¼ significantly improves; , ¼ mixed results thanks to the wider choice in the assignment of tasks to stations. This also confirms previous literature findings. The improvement is more marked when the number of tasks in the assembly sequence increases because this increases also the upper bound DOF max. However, it was noticed that while performance improvement is marked when passing from DOF min to 0.5 DOF max, it becomes negligible passing from 0.5 DOF max to DOF max. This diminishing returns result is similar to that noticed by Portioli (1999). As an example Figures 6 and 7 show a sample trend of h and SR when the average number of DOF increases from the minimum to the maximum value. No generalized conclusion can be given instead for the balance efficiency. In fact BE is more stable, respect variations of TMAX/TC, when both n and TAVG/TMAX are high, while it gets unstable when TAVG/TMAX is low and n is high. This happens because with increasing n a greater number of stations will be required and the unbalance of single stations gets more easily masked when computing the average balance efficiency of the line. With reference to DOF, a higher DOF can even be counterproductive for BE. In fact, with increasing choices it can be easier to balance properly one station and leave other stations poorly balanced. The possibility of having good efficiency at single stations but poor overall line balancing is enhanced in case of low values of TAVG/TC, and TAVG/TMAX (i.e. many short duration tasks easily assignable when DOF is high). This unstable behaviour is especially noticeable when TMAX/TC is about 0.4. As an example Figures 8 and 9 show BE trends vs DOF. Obtained results are summarized in Table III. Product design on assembly lines 1.00 Efficiency 0.90 TMAX = 0.4 Tc 0.80 737 TMAX = 0.5 Tc TMAX = 0.6 Tc 0.70 TMAX = 0.8 Tc N = 30 TAVG 0.4 TMAX TMAX = 0.9 Tc TMAX = Tc 0.60 Min 0.25 Max 0.5 Max Max DOF Figure 6. Effect of DOF on line efficiency and SR 1.60 TMAX = 0.4 Tc 1.50 N = 30 TAVG 0.4 TMAX 1.40 TMAX = 0.5 Tc TMAX = 0.6 Tc TMAX = 0.8 Tc TMAX = 0.9 Tc SR 1.30 TMAX = Tc 1.20 1.10 1.00 0.90 Min 0.25 Max 0.5 Max DOF Max 4.3 Effect of TAVG/ TMAX This parameter is representative of the distribution of tasks duration. High values of TAVG/TMAX indicate a high frequency of long duration tasks and comparatively few short duration tasks. Simulation results show that line efficiency increases with a decreasing TAVG/TMAX ratio, especially when n and DOF are high (i.e. it is preferable to have many short duration operations but with a high DOF in order to saturate easily the workstations) as expected from previously reported literature findings. When instead the assembly sequence is relatively inflexible (DOF is low) it is preferable to have a high TAVG/TMAX ratio (i.e. similar operations mostly of high duration) as already suggested by Kilbridge and Wester (1961). In fact in this case the presence of short tasks but with strong sequence restrictions frequently yields poorly saturated stations because assigning short and long tasks in a forced order may lead to exceed the TC constraint, while if tasks are long then the assignment sequence is less important. Figure 7. Effect of DOF on line efficiency and SR IMDS 108,6 1.00 0.90 0.80 BE 738 TMAX = 0.4 Tc 0.70 TMAX = 0.5 Tc TMAX = 0.6 Tc TMAX = 0.8 Tc 0.60 N = 20 TAVG 0.4 TMAX TMAX = 0.9 Tc Figure 8. Effects of degrees of freedom on balance efficiency TMAX = Tc 0.50 Min 0.25 Max 0.5 Max Max DOF 1.00 0.90 BE 0.80 TMAX = 0.4 Tc TMAX = 0.5 Tc 0.70 TMAX = 0.6 Tc TMAX = 0.8 Tc 0.60 Figure 9. Effects of degrees of freedom on balance efficiency TMAX = Tc 0.50 Min 0.25 Max 0.5 Max Max DOF N Table III. Overview of performance results when DOF number increases N = 30 TAVG 0.8 TMAX TMAX = 0.9 Tc Line Efficiency h Balance efficiency BE Station ratio SR Low High " , " " " , " " TAVG/TMAX Low High , , , , , , TMAX/TC Low High , , , , , , Notes: " ¼ improves; " " ¼ significantly improves; , ¼ mixed results Figure 10 shows sample line efficiency variations when the TAVG/TMAX ratio is increased. In average the balance efficiency was found to improve with increasing values of TAVG/TMAX ratio but there appears to be poor correlation between such two parameters. DOF = Min 1.00 DOF = 0.25 MAX DOF = 0.5MAX Efficiency 0.90 Product design on assembly lines DOF= MAX 739 0.80 0.70 N = 30 TMAX 0.9 Tc 0.60 0.4 0.6 0.8 TAVG / TMAX Figure 10. Effect of average task duration on line efficiency When TAVG/TMAX is high BE is in general less influenced by the number of DOF, while if TAVG/TMAX is low then BE is affected by DOF especially when TMAX/TC is low. In such cases a high DOF may be counterproductive as discussed above. Passing to SR a similar behaviour has been observed, fairly irrespective of the value of n and DOF, when TAVG/TMAX increases. When TMAX/TC ¼ 0.4, then SR is rather stable and gets the lowest (i.e. better) values. When TMAX/TC ¼ 0.5, SR has a decreasing trend while if TMAX/TC ¼ 0.6(0.8 the trend reverses and becomes increasing. When TMAX/TC ¼ 0.9 the trend stabilizes and becomes decreasing for TMAX/TC ¼ 1. In all cases the greater the DOF the better. However, in general, better SR values are obtained when TAVG/TMAX is small. As an example Figures 11 and 12 show opposing SR trends for different values of TMAX/TC ratios. Overall the influence of TAVG/TMAX is questionable and may have mixed effects on the assembly line performance measures. However, it may be in most cases preferable to have low values of TAVG/TMAX ratio. A synthesis of the obtained results is shown in Table IV. 1.60 DOF = Min 1.50 DOF = 0.25 MAX 1.40 DOF = 0.5MAX DOF= MAX SR 1.30 1.20 1.10 N = 20 TMAX 0.5 Tc 1.00 0.90 0.4 0.6 TAVG / TMAX 0.8 Figure 11. Effect of average task duration on SR IMDS 108,6 1.60 DOF = Min 1.50 DOF = 0.25 MAX DOF = 0.5MAX 1.40 SR 740 DOF= MAX 1.30 1.20 1.10 N = 20 TMAX 0.8 Tc 1.00 Figure 12. Effect of average task duration on SR 0.90 0.4 0.6 TAVG / TMAX N Table IV. Overview of performance results when TAVG/TMAX increases Line efficiency h Balance efficiency BE Station ratio SR 0.8 DOF Low High Low High # " , # # " , " " , # # " , TMAX/TC Low High , " , , " # Notes: " ¼ improves; # ¼ worsens; # # ¼ significantly worsens; , ¼ mixed results 4.4 Effect of TMAX/ TC This parameter is affected by the product characteristics and the productivity goal of the line. Therefore, it is representative of a specific line balancing instance. Usually the product designer does not know the value of TC and, for a given product, an optimal TC may exist or not. When line efficiency is considered, better performances are generally obtained for the extreme values of TMAX/TC (i.e. 0.4 and 1) while h worsens for intermediate values of TMAX/TC. In the first case this happens because being TMAX, and therefore TAVG, small respect TC it is easier to saturate stations given the small duration tasks, while in the latter case at least one station can achieve 100 per cent efficiency. This effect is shown for example in Figure 13. Passing to the balance efficiency, BE generally decreases when TMAX/TC increases in cases of low to medium DOF. When medium-high DOF occur an oscillation of BE may occur with wider amplitude especially with the higher DOF and the lower TAVG/TMAX ratio. This effect is shown for example in Figures 14 and 15. As far as the SR ratio is concerned, results depend strongly from the value of the TAVG/TMAX ratio. Often the lowest SR is obtained for the minimum value of TMAX/TC. Higher SR values are obtained with TMAX/TC . 0.4 but non monotonously increasing. In fact, for the cases of TAVG/TMAX ¼ 0.4 and 0.6 a peak of SR is observed at TMAX/TC ¼ 0.5, while for TAVG/TMAX ¼ 0.8 the peak is observed at values of 1.00 TAVG = 0.4 TMAX TAVG = 0.6 TMAX TAVG = 0.8 TMAX Efficiency 0.90 Product design on assembly lines 741 0.80 N = 30 0.70 DOF = 0.25 MAX 0.60 0.4 0.5 0.6 0.7 0.8 0.9 1.0 TMAX / Tc 1.00 Figure 13. Effect of maximum task duration on line efficiency TAVG = 0.4 TMAX TAVG = 0.6 TMAX TAVG = 0.8 TMAX 0.90 BE 0.80 0.70 N = 30 DOF = Min 0.60 0.50 0.4 0.5 0.6 0.7 0.8 0.9 1.0 TMAX / Tc TMAX/TC of 0.6 or 0.8. SR amplitude variations decrease with increasing n but the DOF value seems not significant. As a general rule TMAX/TC should be minimized but the value TMAX/TC ¼ 0.5 should be avoided. As an example Figures 16 and 17 show SR values as a function of TMAX/TC when different values of average DOF and number of tasks apply. As a general rule small TMAX/TC values improve line performances especially when DOF is low. A synthesis of the obtained results is given in Table V Overall the above results confirm the general trends that can be ascertained by the existing literature (Kilbridge and Wester, 1961; Mastor, 1970; Portioli, 1999; Talbot et al., 1986), namely that: . as the mean and maximum task time approach the cycle time the line balancing becomes more difficult (i.e. TMAX/TC should be kept low); . for a given cycle time, the larger the number of small elements available and the fewer the large elements, the easier line balancing becomes (i.e. TAVG/TMAX and Figure 14. Effect of maximum task duration on balance efficiency IMDS 108,6 1.00 0.90 BE 742 0.80 0.70 N = 20 DOF = MAX 0.60 Figure 15. Effect of maximum task duration on balance efficiency 0.50 TAVG = 0.4 TMAX TAVG = 0.6 TMAX TAVG = 0.8 TMAX 0.4 0.5 0.6 0.7 0.8 0.9 1.0 TMAX / Tc 1.50 TAVG = 0.4 TMAX TAVG = 0.6 TMAX TAVG = 0.8 TMAX 1.40 SR 1.30 1.20 1.10 N = 10 1.00 DOF = 0.25 MAX Figure 16. Effect of maximum task duration on SR 0.90 . . . 0.4 0.5 0.6 0.7 TMAX / Tc 0.8 0.9 1.0 TMAX/TC should be kept low and the lower the ratio of long tasks to short tasks the better); irrespective to average task time the presence of small task time elements improves the ability to pack task into stations thus improving balance efficiency; the higher the strength of precedence relations the harder line balancing becomes (i.e. the more parallel is the assembly graph the better is line balancing); and balancing performances improves when number of tasks increases. Nevertheless, detailed findings of this work are not directly comparable with other literature results as this investigation markedly differs from those reported previously. Literature mostly focuses on developing effective balancing algorithms and evaluating their performances instead of systematically evaluating the sensitivity of line performances to parameters variations. For instance, variation of problem instances (i.e. changing number of tasks, order strength and cycle times) only were utilized to 1.50 TAVG = 0.4 TMAX TAVG = 0.6 TMAX TAVG = 0.8 TMAX N = 30 DOF = 0.5 MAX 1.40 Product design on assembly lines SR 1.30 743 1.20 1.10 1.00 0.90 0.4 0.5 0.6 0.7 0.8 0.9 Figure 17. Effect of maximum task duration on SR 1.0 TMAX / Tc N Line efficiency h Balance efficiency BE Station ratio SR DOF Low High Low High , # , , # , , # , , , , TAVG/TMAX Low High , , , , # , Notes: # ¼ worsens; , ¼ mixed results assess the goodness of the utilized balancing technique when determining the minimum number of work stations for a given cycle time or computing the minimum cycle time for a given line length (Mastor, 1970; Talbot et al., 1986). In this work, instead, changes in the assembly problem instances are parametrically investigated in order to assess their impact on line performances, while utilizing the same assembly line balance procedure. 5. Guidelines for designers From the above discussed results some practical guidelines for product designers can be ascertained. 5.1 Average degree of freedom The higher the DOF the better. However, DOF values greater than 0.5 DOF max yield no further increasing benefits in practice. A diminishing returns effect holds as already noticed by Portioli (1999). Therefore, designers should allow at least moderate flexibility in the assembly sequences of the products, but not necessarily strive for excessive flexibility which, anyhow, might not be obtainable in practice. Anyway this appears to be the most sensible parameter. 5.2 TMAX/TC ratio The smaller it is the better. This is to avoid that in case of an inflexible assembly sequence the balance efficiency gets penalized. However, when a suitable degree of Table V. Overview of performance results when TMAX/TC increases IMDS 108,6 freedom exist in the assembly sequence (i.e. DOF . 0.25 DOF max) then the value of TMAX/TC becomes rather ininfluent. This means that designers should aim for short duration tasks maybe splitting longer tasks into smaller subtasks even if this would increase the total tasks number which is contrary to traditional DFMA guidelines. In this case a trade off analysis should be required. 744 5.3 TAVG/TMAX ratio It is difficult to give general guidelines about this parameter. While intuition would suggest that this ratio should be low in general, meaning a high variability of task durations, it was found that its value should be preferably high especially if DOF , 0.5 DOF max, substantially confirming the early empirical observation of Kilbridge and Wester (1961). With higher DOF values the TAVG/TMAX ratio becomes instead scarcely influent on line performances. Only in case TAVG . 0.5 TC (which hardly occurs in practice) it would be preferable to have small values of TAVG/TMAX ratio. This means that designers should aim for a greater number of longer duration task respect shorter duration ones. Even in this case attention is due because increase of task time (maybe through aggregation of subtasks) is against traditional DFMA guidelines and the possible impact of longer assembly time on costs is to be evaluated. 5.4 Number of tasks From the point of view of line performance the higher is the number of tasks the better, because more opportunities for better balancing result. This could be accomplished by splitting tasks in subtasks to avoid increasing total assembly time. It should be noted that increase of task number is opposite to DFMA guidelines which advocate instead task number minimization. In this case a trade off analysis should be required. However, recent literature (Whitney, 2004) confirms that assembly work is no longer considered a major driver of product cost thus relieving criticality of this parameter. 5.5 TAVG/TC ratio This ratio should be kept as small as possible (i.e. increasing TC or reducing TAVG by splitting tasks) so that the probability of finding a short duration task to be allocated to a station with a small residual idle time increases. However, an integer value of (TC/TAVG) £ N should be strictly avoided otherwise a local minimum of performances may occur. Overall such guidelines are aimed at increasing the efficiency of factory operations by improving resource utilization through increase of assembly flexibility. The latter, in fact, contributes to better utilize workstations, avoid bottlenecks, better distribute workloads among stations, reduce the number of workstations for a given production rate or allow to increase productivity for a given line. The increase of DOF besides giving more chances to develop effective and low cost assembly sequences (Whitney, 2004), allows to design reconfigurable assembly lines thus supporting lean manufacturing concepts. It may make easier to switch from manual to automated assembly in case of increased market demand, and enables delayed product differentiation allowing product customization. The greater flexibility in assigning tasks to stations, while preserving line performances, is also useful in case subassemblies are made by several third parties using different facilities and resources. Furthermore, to reduce task durations respect cycle time and to increase the percentage of short duration operations allows the opportunity of reducing cycle time in case it is required to increase throughput and is critical to reduce the defect fractions, which is recognized to be positively correlated to the average task durations (Hinckley, 1993). Finally, the above guidelines do not deal in detail with product configuration, i.e. design of single parts for ease of assembly or simplification of product structure by eliminating unnecessary parts and consolidating multiple parts, which remains the realm of traditional DFMA techniques. 6. Rating index In order to provide product design engineers with an easy to use tool to rapidly assess the goodness of a product design solution with respect to the assembly line performances, a quantitative ranking index is developed here building on previous findings and guidelines. This index can be used to carry out comparative ratings among alternative solutions as well as to supplement traditional DFMA indices. In a previous work (Caputo and Pelagagge, 2006), the above discussed issues prompted to define three indices to assess producibility on an assembly line, namely the station saturation index penalizing the presence of long tasks respect the cycle time: SSI ¼ 1 þ T AVG TC ð10Þ the process time variation index penalising a low variability range of tasks times: PTVI ¼ 1 þ T AVG T MAX ð11Þ and the precedence constraints index which penalizes design solutions characterized by strong precedence constraints among tasks: PCI ¼ 1 þ 1 PI 2 1 ð12Þ where PI is the Portioli’s parallelism index. If PI ¼ 1 then PCI ¼ 3 to avoid an indefinite value. An overall rating index is then computed as: RI ¼ SSI £ PTVI £ PCI ð13Þ This aggregate index has a minimum value of 1 and the higher it gets the worst is the producibility score. However, results from this study enable to gain further insights into this problem and present an improved rating index to be used in alternative or in addition to the previously defined index. The above analysis, in fact, pointed out that the TAVG/TMAX ratio and DOF are the most influencing product-related parameters. The more the DOF approaches the maximum number of DOF corresponding to the specified tasks number the better. Conversely even if some task duration variability is beneficial, an excessive wide range of tasks duration may make difficult to properly balance the line. An intermediate range of Product design on assembly lines 745 IMDS 108,6 746 TAVG/TMAX with a value around 0.6 yields the best results, with higher values less worse than lower values. However, the weight of TAVG/TMAX reduces when DOF increases. Such consideration may be captured by the following ranking index: T AVG T MAX DOFmax 2 DOF RI ¼ 1 þ 2 0:6 DOFmax T MAX T AVG ð14Þ The index has the unit value as a lower bound, meaning that the higher the score the worst is the expected impact of the designer’s choice on the line performance. In this case one obtains the best value RI ¼ 1 when DOF ¼ DOF max or when TAVG/TMAX ¼ 0.6. This index may be also used to complement the ranking index (equation (13)) previously proposed by Caputo and Pelagagge (2006). Such rating indices enable to assess the impact of each design solution on the overall performances of the manufacturing system. The lower the RI, the better a design solution is suited to minimize adverse impacts on the manufacturing system performances. This also enables a rapid relative comparison among competing alternative designs. Finally, the proposed ranking indices may be utilized to supplement the rating obtained from traditional DFMA techniques which usually include an evaluation of the components manufacturing and assembly costs only. In fact, the proposed index focuses on parameters which are undoubtedly relevant, as demonstrated in the present work, but are overlooked by traditional methods which, instead, focus on the minimization of part count and operations cost only without analysing the impact on production system performances. Furthermore, it is also pointed out that when system performance are accounted for, requirements in conflict with those coming from traditional DFMA approaches may result. Therefore, a trade-off decision may arise and the use of complementary ranking methods showing the interacting effects of the parameters may help designers to make an informed choice. 7. Conclusions In this paper a parametric analysis was carried out at first to assess the impact of product features on the overall performances of assembly lines. While the analysis largely confirmed previous literature findings, it also pointed out more subtle interactions between the involved parameters as well as some counter intuitive behaviours. Overall the necessity of striking a balance looking for a trade-off between mutually influencing parameters was highlighted. As a consequence, a rating index was proposed to assess the expected impact of design choices on the management and operation of the manufacturing system. The proposed index appears easy to use, requires only a minimal knowledge of the manufacturing system details, and enables to compare competing design alternatives. This novel DFP methodology, specifically aimed at assembly lines, can supplement the traditional DFMA techniques which are focused on the minimization of components manufacturing and assembly costs but not at optimising the interactions between product design and production system. Adopting the proposed methodology a product designer will be able to rate his design choices and incorporate operational efficiency criteria in the definition of assembly sequences or in the tasks design phase. To this aim some guidelines for designers were also developed to summarize the analysis results. However, the reader is advised that the described results as well as the proposed rating index are based on the analysis of a somewhat limited number of case studies and that generalization of results should be made cautiously. In fact the analysis also showed that complex interactions occurring between the numerous influencing parameters, when facing specific instances of the assembly problem, may significantly affect the assembly line performances in a hardly predictable manner. Therefore, additional research will be needed to analyze in greater detail the interactions between product design features and performances of manufacturing systems in order to derive more general results. In particular, future research work will be aimed at first at extending the parametric analysis (i.e. investigating more complex problem instances with a larger number of tasks, and evaluating in greater detail the role of DOF and task times ratios). A quantitative correlation between line performances and design features will be also searched resorting to response surface methodology. Furthermore, the effect of additional product features will be investigated and other performance measures will be considered. Finally, the role of those design features giving rise to trade off situations will be analysed in greater detail. References Boothroyd, G., Dewhurst, P. and Knight, W. (1994), Product Design for Manufacture and Assembly, Marcel Dekker, Inc., New York, NY. Bralla, J.G. (Ed.) (1999), Design for Manufacturability Handbook, McGraw-Hill, New York, NY. Bramall, D.G., McKay, K.R., Rogers, B.C., Chapman, P., Cheung, W.M. and Maropoulos, P.G. (2003), “Manufacturability analysis of early product designs”, International Journal of Computer Integrated Manufacturing, Vol. 16 Nos 7/8, pp. 501-8. Caputo, A.C. and Pelagagge, P.M. (2006), “Integrating production systems performances in the DFMA methodology”, in Morgan, M.N. and Jenkinson, I.D. (Eds), Proc. 4th International Conference on Manufacturing Research, September 4-7, Liverpool, John Moores University, Advances in Manufacturing Technology. Corti, D. and Portioli-Staudacher, A. (2004), “A concurrent engineering approach to selective implementation of alternative processes”, Robotics & Computer-Integrated Manufacturing, Vol. 20 No. 4, pp. 265-80. Dar-El, E.M. (1973), “MALB – a heuristic technique for balancing large single-model assembly lines”, AIIE Transactions, Vol. 5 No. 4, pp. 343-56. Drejer, A. (2008), “Are you innovative enough?”, International Journal of Innovation and Learning, Vol. 5 No. 1, pp. 1-17. Driscoll, J. and Thilakawardana, D. (2001), “The definition of assembly line balancing difficulty and evaluation of balance solution quality”, Robotics & Computer-Integrated Manufacturing, Vol. 17 Nos 1/2, pp. 81-6. Govil, M.K. and Magrab, E.B. (2000), “Incorporating production concerns in conceptual product design”, International Journal of Production Research, Vol. 38 No. 16, pp. 3823-43. Gupta, S.K., Das, D., Regli, W.C. and Nau, D.S. (1997), “Automated manufacturability analysis: a survey”, Research in Engineering Design, Vol. 9 No. 3, pp. 168-90. Product design on assembly lines 747 IMDS 108,6 748 Hanninen, S. and Kauranen, I. (2007), “Product innovation as micro-strategy”, International Journal of Innovation and Learning, Vol. 4 No. 4, pp. 425-43. Helgeson, W.B. and Birnie, D.P. (1961), “Assembly line balancing using the ranked positional weight technique”, Journal of Industrial Engineering, Vol. 12, pp. 394-8. Hinckley, M. (1993), “A global conformance quality model. A new strategic tool for minimizing defects caused by variation, error, and complexity”, PhD dissertation, Stanford University, Palo Alto, CA. Hundal, M.S. (1997), Systematic Mechanical Designing: A Cost and Management Perspective, ASME Press, New York, NY. Johnson, R.V. (1988), “Optimally balancing large assembly lines with FABLE”, Management Science, Vol. 34 No. 2, pp. 240-53. Khan, Z., Bali, R.K. and Wickramasingh, N. (2007), “Identifying the need for world class manufacturing and best practice for SMEs in the UK”, International Journal of Management and Enterprise Development, Vol. 4 No. 4, pp. 428-40. Kilbridge, M. and Wester, L. (1961), “The balance delay problem”, Management Science, Vol. 8, pp. 69-84. Kusiak, A. and He, D.W. (1997), “Design for agile assembly: an operational perspective”, International Journal of Production Research, Vol. 35 No. 1, pp. 157-78. Lee, C.W. (2007), “The innovation and success of consumer electronics using new product development process”, International Journal of Innovation and Learning, Vol. 4 No. 6, pp. 587-611. Mastor, A.A. (1970), “An experimental investigation and comparative evaluation of production line balancing techniques”, Management Science, Vol. 16 No. 11, pp. 728-46. Nevins, J.L. and Whitney, D.E. (1989), Concurrent Design of Products and Processes, McGraw-Hill, New York, NY. Parsaei, H.R. and Sullivan, W.G. (1993), Handbook of Concurrent Engineering, Chapman & Hall, London. Portioli, A. (1999), “A concurrent engineering approach to assembly line balancing”, Proc. International Conference on Industrial Engineering and Production Management, FUCAM, Glasgow, July 12-15, pp. 342-9. Prasad, B. (1996), Concurrent Engineering Fundamentals: Integrated Product and Process Organization, Prentice-Hall, Upper Saddle River, NJ. Scholl, A. (1999), Balancing and Sequencing of Assembly Lines, 2nd ed., Physica-Verlag, Heidelberg. Seetharaman, A., Sreenivasan, J., Bathamenadan, R. and Sudha, R. (2007), “The impact of just-in-time on costing”, International Journal of Management and Enterprise Development, Vol. 4 No. 6, pp. 635-51. Shukor, S.A. and Axinte, D.A. (2007), “Manufacturability analysis systems: Issues and future trends”, International Journal of Production Research, pp. 1-22. Smith, R.P. (1997), “The historical roots of concurrent engineering fundamentals”, IEEE Transactions on Engineering Management, Vol. 44 No. 1, pp. 67-78. Smith, P.G. and Reinertsen, D.G. (1991), Developing Products in Half the Time, Van Nostrand Reinhold, New York, NY. Soundar, P. and Bao, H.P. (1994), “Concurrent design of products for manufacturing system performance”, Proceedings of the 1994 IEEE International Engineering Management Conference, pp. 233-40. Talbot, F.B., Patterson, J.H. and Gehrlein, W.V. (1986), “A comparative evaluation of heuristic line balancing techniques”, Management Science, Vol. 32 No. 4, pp. 430-54. Wee, T.S. and Magazine, M.J. (1981), “An efficient branch and bound algorithm for an assembly line balancing problem. Part II: maximize the production rate”, Working Paper No. 151, University of Waterloo, Waterloo. Whitney, D.E. (2004), Mechanical Assemblies, Oxford University Press, New York, NY. Product design on assembly lines 749 Corresponding author Pacifico M. Pelagagge can be contacted at: [email protected] To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints
© Copyright 2026 Paperzz