• No results found

Assembly Line Balancing Application

4.4 Part Allocation at Workstations

5.1.1 Assembly Line Balancing Application

For the purpose of balancing the assembly line, an application was created by the researcher in Delphi Embarcadero RAD Studio XE7. The application takes into account all requirements and restrictions presented in Section 5.1. It can be used for the three types of assembly line balancing problems that were discussed in Section 3.1.4, being the SALBP-I, SALBP-II and SALBP-E. The application decides for itself which problem it is optimizing based on the variables that can be provided to the application. The interface of the application is presented in Figure 5-1.

For the purpose of this research the program solved SALBP-E problems where both the cycle time and the number of workstations are allowed within certain ranges. The objective of this problem is maximizing the line balancing efficiency. This is the same as minimizing the total time (Gurevsky, BattaΓ―a, & Dolgui, 2013), which we will demonstrate with an example.

Section 3.1.4 presented the following formulas:

1) π‘‡π‘œπ‘‘π‘Žπ‘™ π‘‘π‘–π‘šπ‘’ = π‘π‘Ÿ. π‘œπ‘“ π‘Šπ‘œπ‘Ÿπ‘˜π‘ π‘‘π‘Žπ‘‘π‘–π‘œπ‘›π‘  βˆ— πΏπ‘œπ‘›π‘”π‘’π‘ π‘‘ π‘ π‘‘π‘Žπ‘‘π‘–π‘œπ‘› π‘™π‘œπ‘Žπ‘‘

2) 𝐿𝑖𝑛𝑒 π΅π‘Žπ‘™π‘Žπ‘›π‘π‘–π‘›π‘” 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 (%) = π‘‡π‘œπ‘‘π‘Žπ‘™ π‘‡π‘Žπ‘ π‘˜ π‘‡π‘–π‘šπ‘’ (𝑠𝑒𝑐)

π‘π‘’π‘šπ‘π‘’π‘Ÿ π‘œπ‘“ π‘Šπ‘œπ‘Ÿπ‘˜π‘ π‘‘π‘Žπ‘‘π‘–π‘œπ‘›π‘  βˆ— 𝐢𝑦𝑐𝑙𝑒 π‘‡π‘–π‘šπ‘’ (𝑠𝑒𝑐)βˆ— 100%

A problem with 3 tasks of 25 seconds and 1 task of 20 seconds can for example be divided as follows:

1 2 3 4

The corresponding values for the two formulas are:

Configuration Total Time Line Balancing Efficiency

1 4 * 25 = 100 (3*25 + 20)/(4*25) * 100% = 95%

2 2 * 50 = 100 (3*25 + 20)/(2*50) * 100% = 95%

3 3 * 50 = 150 (3*25 + 20)/(3*50) * 100% = 63.3%

4 2 * 75 = 150 (3*25 + 20)/(2*75) * 100% = 63.3%

Minimization of the first formula results in configuration 1 or 2, maximization of the second formula also results in configuration 1 or 2. Therefore, we maximize the line efficiency if we minimize the total time spent.

39

Figure 5-1 Print Screen of the Line Balancing Application

To use the application, the user has to collect certain operation data which is described in the (Dutch) manual in Appendix 3. If the user has collected all the data, the user can use the application for an SALBP-E problem by giving in a random cycle time that is too high as well as a lower bound. This lower bound is for example the total task time divided by the highest number of workstations possible. Next the user gives in the allowed range for the number of workstations. When clicking on the button β€˜Start Algorithm’, the application starts running from the entered cycle time. For every cycle time until the lower bound, the application creates an initial sequence by scheduling the longest task of the available tasks given the precedence constraints. The application also takes into account process cycles of tests, as it prefers scheduling other tasks over scheduling waiting time. Then the algorithm tries to improve the sequence by inserting tasks on other possible places, hereby, also taking into account the precedence constraints. The objective of this step is to improve the vertical balance. Recall that the Smoothness Index is a measure for the vertical balance (Section 3.1.4). Minimization of this formula maximizes the vertical balance.

40 To not end up in a local optimum, a simulated annealing (SA) algorithm as introduced in Section 3.2 was implemented for this step. The SA code can be found in Appendix 2. Recall that simulated annealing requires a cooling scheme that has four variables: starting temperature, stop temperature, Markov chain length and decreasing factor. The cooling scheme used for this research was determined by structurally evaluating multiple combinations of inputs. The combinations, results and explanation are included in Appendix 2 as well. The user is able to change the variables of the cooling scheme when using other data.

The cooling scheme for the data used in this research is as follows: Starting temperature = 80

Decreasing factor = 0.9

Markov chain length = 10 Stopping temperature = 0.5

When the SA algorithm is finished, the application checks whether the sequence can be performed with a lower cycle time. If this is the case it stores this cycle time as the final cycle time for the sequence. It then determines the corresponding total time required. After every sequence, it is checked whether the total time has improved (decreased). If this is the case, the sequence is stored as the best solution. After a proposed solution is checked for improved performance, the cycle time is decreased by 1 and another sequence is created.

This application can be used whenever changes are made to operations in the assembly line. Whenever a change occurs, data should be updated, the cooling scheme should be adjusted and the application can be run again. The program will then again determine the most efficient task allocation in terms of line balancing efficiency.

Validation of the Line Balancing Application

Validation of the Line Balancing Application is done by loading the exact same task division into the program and comparing the numbers for the Smoothness Index and the Line Balancing Efficiency with the values that were calculated by hand. For this purpose, the task division used at the beginning of this research was used. The program returns the same values for both the Smoothness Index and the Line Balancing Efficiency that were calculated by hand (Figure 5-2). Therefore, it is concluded that the program is able to reflect reality and can be used for determining the consequences of changes. The values calculated by hand were 84.7 for the Smoothness Index and 90.9% for the Line Balancing Efficiency. The small deviations are caused by rounding.

Figure 5-2 Validation of the Line Balancing Application

Baseline result of Line Balancing Application

The application is used to improve the line balance of the current production line. Running the algorithm for the existing operations, precedence constraints, test data and fixed times (walking & stamping) that have to be taken into account, results in the task division as presented in Figure 5-4. Figure 5-3 shows the progress of the application with regard to the Line Balancing Efficiency.

41

Figure 5-3 Progress of Line Balancing Application for the Current Situation

The program finds a near-optimal solution for the task division with 11 workstations and a cycle time of 226 seconds. This is 6 seconds less than the current cycle time of 232 seconds. This means that a total of 66 seconds can be saved on the production of a boiler on this production line. The Line Balancing Efficiency of this task division is 95.2% and the value of the Smoothness Index is 52.6 as presented in Figure 5-3. Recall that the values of the current situation are a line balancing efficiency of 91.4% and a smoothness index of 65.82. This means that the application found an improvement of - 2.6% for the cycle time, +4.2% with regard to the Line Balancing Efficiency and an improvement of - 20.1% for the value of the Smoothness Index. The biggest differences between the old and new task division are the more even balance for workstation 1 to 3, the separation of the beginning and end of the heatcell test over 2 workstations and that workstation 10 prepares the accessory supply for workstation 11. It was expected that the application would find a better result by some small task exchanges, but a cycle time reduction of 6 seconds exceeded expectations.

Figure 5-4 presents the task division as found near-optimal by the line balancing application. In this figure, every bar resembles a workstation on which an employee is stationed. The coloured blocks within the bar resemble different tasks and the length corresponds to the duration of a task. The colours do not have a meaning, this is purely for visualization purposes. The number inside the task corresponds to an operation which is part of the input data. For example, task 52 is scheduled on

workstation 4 and resembles the operation: β€˜Locate and secure transformer bracket to wb7’ with a

duration of 14.4 seconds. The cycle time of the solution is presented next to the indication of the

42

Figure 5-4 Task Division Result of Line Balancing Application applied to Current Situation

Economics

The Line Balancing Application determined a cycle time reduction of 6 seconds. When multiplying this with the forecasted production numbers over the next couple of years and converting this to time savings in Euros, one can calculate the NPV (Recall that the formula is mentioned in the literature review in section 3.4). To implement the changes in the task division, the production line has to be reconfigured by the TEF Department, it is approximated that this costs around 2000 Euros. It is assumed that these changes can be done outside of production hours. If this isn’t the case, the production losses should also be included. To account for inflation, the savings are incremented with 2% every year. The interest rate used within Bosch is obtained from the financial controller. Let us now look at the NPV calculation results (Table 5-1).

43 Year Initial Investment (€) Reconfiguration Costs (€) Cycle Time Savings (€)

Yearly Cash Flow

(€) Cumulative Cash Flow (€) NPV (€) 1 3864 2000 16675 10811 10811 10811 2 15864 15864 26675 25365 3 15386 15386 42060 38315 4 14463 14463 56523 49483

Table 5-1 Economic Evaluation of Implementation of the Line Balancing Application

The pay-back period of the investment is only 129 days. This is well within the allowed pay-back period of 3 years and using the program seems very beneficial.

Ergonomics

This solution does not worsen or improve the ergonomic situation. In total, the exact same operations are performed which means that the total ergonomic score does not change. However, there are small differences across the workstations which can impact the individual ergonomic scores of the workstations. As the changes do not impact the overall score, we consider the implementation of this change to the production line as ergonomically indifferent (0).

Further Use of the Line Balancing Application

Now, we have determined the improved situation of the current situation without any changes to the production line. This improved situation serves as a baseline for the changes that are proposed in the subsequent sections. All changes that seem to be beneficial to the efficiency of the production line are run through the application to see whether it can improve the total time required even further. The results of the Line Balancing Application are then used for economical assessment of the proposals. Let us now have a look at several possibilities for further improvement.

Related documents