Parametric Cost Estimation Model for Li-ion Battery Pack of E-motorcycle Conversion based on Activity Based Costing

: Universitas Sebelas Maret (UNS) through SMART UNS Company has conducted research and development of e-motorcycle conversion using Li-ion battery pack as a substitute for ICE energy source from the conventional motorcycle. Currently, the battery-pack that used for e-motorcycle conversion is in the development phase towards commercialization. The challenge of estimating production costs is the complicated production process and storing hidden expenses that can be a problem. This hidden cost is often a missing or varied factor that costs less or more expensive. This study presents an integrated parametric cost estimation model with activity-based cost assignments to estimate production costs through cost calculations for each activity. Activity-based costs break the production process into a specific cost element for each step. Each activity's cost is put into a parametric cost estimation model to calculate the cost of each activity into the total cost of production. Cost estimation results will be analyzed using a regression method to determine which variables most affect the production cost of Li-ion battery packs for the conversion of e-motorcycles in the SMART UNS company.


Introduction
The level of motorcycle sales in Indonesia in 2019 increased by 1.6% from the same period last year [1]. The motorcycle still uses the Internal Combustion Engine (ICE) technology, which is very influential on the level of fuel oil usage from fossil energy [2]. For ten years, the consumption of fossil fuels increased by an average of 1.3 percent annually [3] and its annual amount of 14% emissions caused by fossil fuels from the transportation sector. Emissions from the transport sector are mainly coming from vehicles that dominate the release of long-lived greenhouse gases. This makes increased contributions to the total effect of the anthropogenic greenhouse [4]. Emissions resulting from fossil fuels cause an increase in CO2 that results in climate change [5]. The growth rate of CO2 has a strong correlation with global temperature anomalies with CO2. Global warming rates have been accelerated in the last decade. The global surface temperature in 2019 is the 2nd highest in the period of instrumental measurement in the Goddard Institute for Space Studies (GISS) analysis. The global temperature 2019 is + 1.2 °c (~ 2.2 °f) warmer than in the base period 1880-1920 is a reasonable estimate of the 'pre-industrial' temperature [6]. Electric vehicles are automotive products that have capabilities to improve vehicle performance and mitigate the negative effects of the environment [7]. Electric vehicles contribute to the reduction of greenhouse gas emissions evidenced by previous research to date has shown that electric vehicles produce lower greenhouse gas emissions [8]- [9].
One alternative offered as an effort to overcome the problem in this is to use electricity technology to be used as an energy source in all types of vehicles. The use of batteries as energy

Data Collection
This study developed a parametric cost estimation model with an activity-based costing approach. Figure 1 shows the flow of research using this approach. The initial phase of this research begins with collecting data such as the bill of materials from battery packs and business processes from the SMART UNS company. The bill of materials and business processes of the Li-ion battery pack by SMART UNS are shown in Figure 2 and Figure 3. Through these data, we can identify cost driver rates and cost centers for each activity. Cost driver rates are the main component in the parametric cost estimation model. The estimated cost estimation model is used to calculate production costs.

Figure 1. Research Process
The development cost estimation model starts with identifying the BOM Li-ion battery pack. The BOM tree structure of a Li-ion battery is shown in Figure 2. Through the bill of material provides information related to the components forming a Li-ion battery pack. BOM is used as an essential data parameter in product life cycle management that represents product information such as the hierarchical part associated with a particular product. Through multi-levels BOM can be used to determine the engineering bill of materials needed in estimating costs.

Figure 2. Bill of Material
The business process at SMART UNS is divided into 4 groups, such as management as administration and activities outside of production, the Li-ion battery module team is a group working on a production at the battery module stage. The electrical component assembly team is a group working on the production process of the electrical component assembly stage. The charging and testing team is a group that is working on the final stages of the production process, namely charging and final tests. There are two types of li-ion battery packs produced at SMART UNS type A for 150cc and type B for 110cc. Both models have the same production process, and the difference is the specifications of the material used.

Cost Driver and Cost Center Identification
Activity-based costs are developed to get a more accurate cost estimate [23], [24]. The difference between activity-based costing with the traditional system is the determination of cost drivers [25]- [27]. Cost driver is a driving factor that triggers cost and intermediate factors between cost objects with activities and resources [28]. For selecting the cost driver should be done carefully to ensure the accuracy of the cost. Some researchers have conducted research related to cost drivers as conducted by Cokins and Căpuşneanu [27], Sheng [28], Geiger [29], Răvaş and Monea [30], Dražić-Lutilsky and Dragija [31].
According to Sheng, the cost driver has some specific characteristics such as concealment, relevance, application, and accountability [28]. Cost drivers must have a causality relationship with the activities and costs, it must be measured and explain the use of resources consumed during an activity [30]. Cost drivers should demonstrate correctly the relationship between specific activity and cost objects [31], [32]. One of the requirements of the construction of the cost driver is the cost parameter. Each cost driver relates directly to the process engineering, it can be used in creating task chains. The engineering process associated with this cost driver can be triggered to generate value for cost parameters.
SMART UNS with the business processes that have been described are important for identifying costs associated with various activities in the process, and this is to assess and evaluate inefficiencies based on their economic impacts. The activity-based costing approach in the parametric cost estimation model begins with defining general activities and their cost drivers. Activities, cost drivers, and cost centers can be identified through the results of field observations and maps of the operation process of producing Li-ion battery packs for e-motorcycle conversion. Some studies are used as a reference in the determination of cost driver and parameter costs, including research conducted by Sutopo, Nizam, Purwanto, Atikah and Putri [20]; W. Sutopo; A. Eliza; R. Ardiansyah; Yuniaristanto; and M. Nizam. Parametric [21], MY Abu; KR Jamaludin; and MA Zakaria [22], Fog [33], Erick Ten Bright [34], Katrin and Tatjana [35].

Parametric Cost Estimation Model Development
In calculating the cost estimation model with the activity-based costing approach, it can generally be done by multiplying the cost driver rate by the number of driving activities as in equation (1).The cost driver rate is the cost that must be incurred for each activity undertaken. The equation (2) is used to calculate cost driver rates.
Where is activity costs j, is cost driver k for activity j, is number of activity in activity, is total activity costs j, is the predicts number of activity drivers k in activity j. The total cost of the activity considers several things. In indirect activities, it is necessary to consider overhead costs such as indirect labor costs, machine depreciation costs, electricity costs, consumables costs, and other costs that support these activities. Whereas the direct activity costs that are considered in the calculation of activity costs are the direct labor cost and raw material costs.

Monte Carlo Simulation
In this research, monte Carlo simulations are carried out to produce data on the amount of production, considering that the battery pack is a new product with no historical data. The monte Carlo simulation through the generation of random numbers is performed using the function on equation (3). This data is used in multiple linear regression analysis to analyze the variables that affect the cost of making a battery pack at SMART UNS.

Parametric Cost Estimation Model
Through business processes and field observations, it was found that there were ten indirect activities and 11 direct activities. Through these activities, the cost driver is identified to build a parametric cost estimation model. Table 2 shows the parametric cost estimation model with an activity-based costing approach. Total Production Type B

Numerical Example
In this section, an estimated battery-pack production cost is calculated for one period. This section begins with calculating the cost driver rates for each activity using equation two and calculates the production cost using the parametric cost estimation model in table 2. Table 3 is a recapitulation of the calculation of cost driver rates for each activity. The calculation uses the parametric cost estimation model in table 2 for a period of 1 month with total production for type A is 40 units and type B for 35 units. Table 4 is calculated the estimated costs for total production and unit costs of each type of li-ion battery pack for e-motorcycle conversion. Detailed calculations of estimated production costs for one month are shown in Appendix A2.

Simulation Design
In this section, the Monte Carlo simulation design for the number of lithium-ion battery pack production for e-motorcycle conversion. Monte Carlo simulation aims to develop data that will be used to analyze multiple linear regression models. Monte Carlo simulations can predict errors from simulations that are proportional to the number of iterations. For new products, the specified error value is 58% [36]. Equation (4) to calculate the number of iterations needed to get a result with an error of 58%.
Where is the number of iterations, is a standard deviation, and is an error value. The results of the calculation of the number of repetitions using equation (4) are 1512 iterations.

Estimation of Multiple Linear Regression Models
This section analyzes multiple linear regression to establish the relationship between the dependent and independent variables. There are 3 regression models built, the first regression model to determine the total cost of producing lithium-ion battery packs for e-motorcycle conversions. The second and third regression models are used to determine the cost of production per unit of lithiumion battery packs for Type A and Type B. In calculating cost estimation using activity-based costing, costs are triggered by the existence of resource usage activity. Each activity has an activity cost driver that determines the number of costs incurred according to the resources used. Wagner (2012) stating that production volumes are a fundamental trigger cost. Therefore, independent variables in multiple linear regression analyses used the number of total production and the number of battery-pack production for type A and type B. This section is used IBM SPSS Statistics 25 software to estimate the regression model between the dependent variable and the independent variable.
: Unit production cost of Lithium-ion pack-battery for Type B (USD) 1 : Total production of lithium-ion battery packs for Type A (Unit) 2 : Total production of lithium-ion battery packs (Unit)

Classical Assumption Test
The classic assumption test aims to provide certainty that the regression equation obtained has accuracy in estimation, is unbiased and consistent. The classical assumption test consists of 4 parts, namely, multicollinearity, autocorrelation, heteroscedasticity, and normality.   Figure 3 it is known that the distribution of points is relatively close to the diagonal line so that the residual data criteria are normally distributed with the Normal Plot approach.

Model Feasibility Test
In this section, test the estimation of the regression model that has been formed in section 3.4 to measure the accuracy of the regression model in estimating the actual value. This section uses the F test and the T test.

F Test
In this section, an F test is performed to determine whether the independent variables simultaneously affect the dependent variable. 2 hypotheses are used. In general, these two hypotheses are: a. H0 = Simultaneous independent variables do not significantly influence the dependent variable. b. H1 = Simultaneous independent variables simultaneously have a significant effect on the dependent variable. Hypothesis testing is done by comparing the significance value with 5%. If the significance value < 0.05 then H0 is rejected, and if the significance value> 0.05 then H0 is accepted. Because the significance value <0.05, the three models have a simultaneous influence between the independent variables on the dependent variable.

T Test
T-test is a method of testing the model to determine the effect of each regression coefficient on the dependent variable. There are 2 hypotheses are used, in general these two hypotheses are: a. H0 = The independent variable has no significant effect on the dependent variable. b. H1 = The independent variable has a significant effect on the dependent variable. Hypothesis testing is done by comparing the significance value with 5%. If the significance value <0.05 then H0 is rejected, and if the significance value> 0.05 then H0 is accepted.  10 it is known that in 1st model the independent variables have a significant effect on the dependent variable. In 2nd model it is known that the total variable production of battery type A pack does not have a significant effect on the dependent variable, while the total production variable has a significant effect on the dependent variable. Then the 3rd model it is known that the total variable production of battery type A pack does not have a significant effect on the dependent variable, while the total production variable has a significant effect on the dependent variable. Although there are non-significant variables in the second and third models, the model can still be used because if the model runs simultaneously, significant variables will influence the insignificant variables.

Determination of the Most Influential Variables
To find out the independent variables that most influence the dependent variable, use the Standard Coefficient Beta test. The highest beta coefficient marks the independent variable that has the biggest effect. Total production of battery-packs Type A .002 X Total production of battery packs -.997 V a. Dependent Variable: Unit production cost of pack-battery Type B The relation of total cost per unit with change of the activity output is known by reviewing the behavior of cost [37]. In literature, the cost behavior is described as fixed or variable with respect to changes in production volumes. Volumes of output as the fundamental cost driver. Variable costs change proportionally to the change in production volumes [38]- [39]. In standard cost models, variable costs change proportionately with changes in the activity driver, implying that the magnitude of a change in costs depends only on the extent of a change in the level of activity, not on the direction of the change [40].
Based on the first model, it is known that the number of production variables has the highest beta coefficient, 0.917. Therefore, production costs are more influenced by the number of production variables compared to other variables. The factor owned by the total variable production is positive, this shows that if the total production increases, the total production cost will increase as well. Based on the second model, it is known that the number of production variables has the highest beta coefficient, which is 0.997. Therefore, production costs are more influenced by the number of production variables compared to other variables. The coefficient owned by the total production variable is negative, this shows that if the total production increases, the production cost of type A battery units will be smaller. Based on the third model it is known that the number of production variables has the highest beta coefficient value, which is 0.997. Therefore, production costs are more influenced by the number of production variables compared to other variables. The coefficient owned by the total production variable is negative, this shows that if the total production increases, the cost of producing a B type battery unit will be smaller. In this study when the assumption expanded by enlarging the value of significance then it is possible that the H0 can be accepted even if it is wrong and resulting in a change of influence between the dependent variables to the independent variables.
Based on the analysis, the company can maximize the amount of production according to the production capacity to reduce the cost of unit production. By maximizing the amount of production, the price of the product can be more competitive. To achieve production capacity, the company can create an operation process chart and apply standard operational procedures without override product quality.

Conclusions
This research chooses an activity-based costing method to classify the cost of producing a Li-ion battery pack for e-motorcycle conversion. Activity-based costing methods provide a more accurate view of product costs than traditional cost methods by identifying each activity element's entire cost. This method determines all activities related to the production process, allocates costs for these activities, and helps classify the production process costs more easily and faster.
The activity-based costing method is integrated with the parametric cost estimation method, which is the right method to be applied with the estimated cost of producing a Li-ion battery-pack for e-motorcycle conversion through a mathematical model. Cost estimation results that reflect a significant difference between product specifications. In addition, cost estimation also reflects the overall use of the company's resources. Activity-based costing helps companies in resource management to get a more competitive cost. Moreover, changes in the operation process for cost reduction will allow the company to fulfill customer needs. Therefore, the battery-pack company can use the activity-based costing method to accurately estimate the cost.
This research also used the regression analysis to analyze. The results of data processing show that the total production costs and unit production costs of the two types have the greatest influence on the total production. However, the total production variable has a different effect on the calculation of total production costs and unit production costs. In the calculation of the total production costs, if the amount of production is increases, the total production costs will increase. Whereas in the calculation of unit production costs it is known that if the total production amount increases, the unit production costs will decrease.    Hourly quality control and inspection rates (USD / hour)