Development of Variable-Line Balancing Chart by Risk Assessment Using Monte Carlo Simulation

Line Balancing Chart is one of the most used lean tools for determining cycle time (CT) and lead time (LT) in the production line processes flow, from raw material input until finished product output. However, for traditional Line Balancing Chart, only the average (mean) CT and LT are considered which causing inconsistent performance of the actual production line throughput rate (productivity). In this study, Variable-Line Balancing Chart (V-LBC) is introduced by considering the dynamic CT and LT in a form of (Minimum, Most Likely (mean), Maximum) for each process involved in the production line. The risk assessment for Value-added (VAA) and non-value added (NVAA) events in the flow are also considered for this V-LBC. A Monte Carlo simulation by using @Risk software is utilized to simulate each process CT distribution capability. As a result, each process in the V-LBC could be represented in horizontal and vertical time variables that involve a variable CT (VAA and NVAA) and risk assessment using Risk Assessment-Failure Mode and Effect Analysis (RA-FMEA) approach. The actual root cause led to the process variation also could be identified more accurately from the V-LBC. Hence the correct action could be taken in order to reduce the variation which indirectly increase the production line productivity.


Introduction
The 100% efficiency production line is an ultimate target for all manufacturing industries.This could be achieved if the management strictly implements lean and six sigma culture in the company.Besides, all the processes tact time at the production line must be balanced in order to eliminate production wastes such as waiting time, over production, defects and movement wastes [3].Various lean and six sigma tools could be selected as guidelines for actions taken either by reducing the process variations (improve six sigma level) or by eliminating the production wastes from the production line [4] [5].Line Balancing Chart (LBC) is one of the best methods that could be applied for analysing the processes' tact time at the production line.Even though the analysis was done by using traditional LBC, the 100% efficiency is still difficult to be achieved.One main limitation of the traditional LBC is it only considers the average tact time value for each process.In fact, there are hidden times which causing the variation of the processes including production wastes and risks [1] [2].
The tact time at each process normally will consists of value-added activities (VAA) and non-valueadded activities (NVAA).In addition, there is an unexpected time added due to the unplanned risk happen during the process ongoing [1].Therefore the variable line balancing chart (V-LBC) is introduced to consider all VAA, NVAA and risks attributes in analysing the process tact time at each workstation.

Figure 1.
Production line layout for 3-in-1 cocoa powder drink at company M Company M's production consists of six operators located at six processes to complete the whole cycle of the completed mixed 3-in-1 cocoa product.The process includes mixer, filling gross weight, filling fine weight, sealing, visual check and packing/batch marking.It begins with collecting Tact Time (TT) data for each process involve in producing the product.For each process, 50 data were recorded as an input data by recorded video analysis.

Methodology
The development of V-LBC model consists of three (3) main phases; Define, Measure and Analyse.Figure 2 shows the methodology flow of this model.In Define phase, the flow starts with a selection of product family.All types of products were listed out and a process flow to complete one cycle of each product is studied.Products with the most common process flow will be selected as this will give a bigger impact in term of time and cost improvement once the action carried out.Besides, the selection is also need to base on the company's CEO decision since it might give impact on the overall profit of the company, for instance the product running under OEM product is good to be selected due to meeting the customer satisfaction in term of quality and quantity.Next, the products which are called scenario in this study, will be selected for further study on the production line balancing and RA.

Figure 2. V-LBC Model
In Measure phase, it begins with collecting Tact Time (TT) data for each process involve in the scenario.For each process, 50 data need to be recorded as an input data.The researcher's team need to go to the production line to view (gemba) by himself the actual processes happen in the line.One the tact time data collection completed, a brainstorming session was set up together with the industry's team to identify production wastes, work in progress (WIP) and potential risks to occur at each process.Then, in Analyse phase, all the data measured are divided into two (2) categories : i) Category 1: Value added activities (VAA) data from processes' tact time and Non-value added activities (NVAA) data from production wastes identified from brainstorming session.ii) Category 2 : Risk data from brainstorming session All data are recorded and converted in (Min, M/L, Max) seconds.For category 1, VAA and NVAA data for all six processes are tabulated.These input data then fitted by using @Risk software to identify the best type of distribution to represent the data population.This data fitting is necessary for a preparation of simulation process by using Monte Carlo simulation in the following process.Once the data is fitted, VAA/ NVAA input table is ready as shown in Table 1.As for category 2 data, type of possible risks for each process to be recorded and tabulated.By using Risk Assessment-Failure Mode & Effect Analysis (RA-FMEA) approach, all the risks to be ranked follow its severity, occurrence and detectability level.Then they are converted into (Min, M/L, Max) time as well, as agreed during brainstorming session between researcher's and management teams.Risks data also need to be fitted by @Risk software before summarized in a Risk Registered table shown in Table 2. Table 2. Risk Registered Input Table

Monte Carlo Simulation
The Monte Carlo simulation is run by using @Risk version 8.1 software.Two worksheets consists of CT/LT data (min, M/L, max) with probability distribution defined (Table 1) and Risk Register (Table 2) will be used as input to the Monte Carlo simulation.By using this simulation, the result gained could be more accurate since different input data within the 90% confidence interval will be repeated 1000 times besides 20 times simulation.This may save time consumption compare to traditional method calculation.The output for overall simulations will be (min, M/L, max) values of total CT, total LT, total time added to plan (due to risk) and total cost added to plan (due to risk).The result for each parameter to be compared with the traditional line balancing chart model in order to determine the impact of variability and risk to the mean value.Finally, the current state variable -line balancing chart (V-LBC) model to be drawn by using the input and output data from the Monte Carlo simulation.

Results and Discussion
From the simulated VAA, NVAA and risk data obtained, the Variable-Line Balancing Chart (V-LBC) could be plotted.The major different between V-LBC compared to a traditional line balancing chart is by considering all actual variations occur in the production line's processes which includes tact time (VAA), production waste time (NVAA) and time added to plan due to process risks (RA time added to plan).The overall V-LBC is as shown in Figure 3.

Min Most Likely Max Mean
Time Added to Plan (sec)

Most Likely Max Mean
Cost Added to Plan (RM)  For simulation use : define data (risk cost) distribution type

Figure 3. V-LBC for company M production
Each process's TT (process-1 until 6) need to be plotted by considering the (min, M/L, max) time, which is obtained from the 50 data recorded.This type of variable plot is call horizontal variable since it is plotted in horizontal manner.Then, each variable plot will be plotted its tact time gained from the Monte Carlo simulations which consists of VAA time (CT), NVAA time (production waste) and Time added to plan (RA time).These plots that gained from the simulations are called vertical variable and plotted in vertical direction.Besides that, number of work in progress, WIP (converted into equivalent WIP time) could also be monitored since it is included in the NVAA time calculation.
For company M production line, the horizontal variable shows a big variation (min, M/L,max) at process 1-Mixer, 2-Filling Gross Weight and 5-Filling Visual Check.This process variation could be due to unevenness (Muda) in the process.Standardization on the process sequencing and control chart are few six-sigma tools which could be used to overcome this issue.Kaizen activities required to be carried out in order to reduce the variation, hence meeting the Cycle Time (CT) line of the V-LBC for all (min,M/L,max) time.For vertical variable, three columns bar chart represents the after-simulation results from VAA/ NVAA value (using VAA/NVAA input -Table 1) and Risk value (using Risk Register input -Table 2).The contribution of each factor (VAA, NVAA or risk) could be easily identified from the V-LBC.For instance, from Figure 3, at process 1-Mixer, maximum tact time big variation is dominantly affected by risk factor.Actions could be focused on how to eliminate the risk at this process.As for process 5-Filling Visual Check, NVAA factor seems a dominant factor regardless at minimum, most likely (average) or maximum tact time variation.Therefore, by focussing at the NVAA factor, mainly due to production wastes which already determined for this process earlier, may stabilize the variation at this process.Besides, the abnormality of WIP distribution along the production line could also be known from V-LBC chart.In Figure 3, abnormality of WIP at mixer station is due to the nature of mixer process that need to run in batches.Therefore, it is required to management to plan the batches carefully to avoid the production downtime at a same time optimize the inventory level at the production line.

Conclusion
This project aimed to develop a Variable-Line Balancing Chart model.The V-LBC plot which consists of horizontal and vertical variables are believed to be more realistic tact time to be considered at each work station for a better countermeasure for productivity improvement.The tact time in horizontal and vertical variable are more wholistic since besides it considers the variation in process CT, the NVAA and RA time are also considered as well compared to a traditional LBC which focuses only on the mean value of CT.Finally, the management could view the real condition of each process in the production line including the Balancing Chart for Scenario 1 (Munif Cocoa 1kg pouch bag) + RA RA-Time Added to plan (s) NVAA Waste (s) VAA Tact time (s) CT = 19.5 s (1200 pack) CT = 14.6 s (1600 pack) CT = 13 s (1800 pack) NVAA No. of WIP