### Introduction

### Materials and Methods

### 1. Learning-Forgetting Curve Model

*t*= Time required to produce x units in cycle_{xi}*i*(*hr*),*T*_{1}= Time required to produce the first unit (*hr*),*x*= Number of units produced in cycle_{i}*i*,*u*= Value of experience remembered at the start of cycle_{i}*i*,*b*= Learning index,*f*= Forgetting index,_{i}*S*=Total output when there is no break,_{i}*τ*=Length of break (_{i}*hr*),*D*=Time required to completely forget a task (*hr*),*LR*=Learning rate.

*u*was defined as the level of memory that the worker remembers about the operation in the rest phase after the operation phase.

*LR*value was reported to be about 80%, on average. Because this value is predicted to be consistent with the median value of the

*LR*range of a similar construction industry, the

*LR*value was assumed to be 0.8 [21].

*x*

*is one unit of work, and*

_{i}*u*

*is an indicator of how much a worker remembers about a task when he has taken a break. This value increases proportionally to how much work was performed previously and how much was previously remembered.*

_{i}*D*, the time required for the worker to completely forget the task, was assumed to be 6 months used in related studies [22].

*x*

*is multiplied by*

_{i}*tx*

*, the working time*

_{i}*t*is calculated. Equation (1) can then be expressed as Equation (7).

*x*

*, according to the working time*

_{i}*t*, can be calculated using Equation (7). The changes in other factors, according to the working time

*t*, are also calculated. When

*t*=0, the initial conditions of

*x*

_{0}and

*u*

_{0}are 0 and 0, respectively, assuming that both the initial work produced and the learning about the task start from 0.

### 2. Scenario Assumptions

*T*

_{1}must be derived for the LFCM correction of a trained but inexperienced (basic-trained) worker. A skilled worker is assumed to be sufficiently trained and experienced; therefore, it is considered that there is little change in

*tx*

*with time.*

_{i}*T*

_{1}shown in Equation (7) is the speed at which a basic-trained person performs a task. In practice, it is difficult to know the exact

*T*

_{1}because the decommissioning task is being performed by a skilled worker. In previous research, the

*tx*

*of a skilled worker was calculated as 3.06 hours for decommissioning a bio-shield. A correction based on this value was required. To determine the initial value*

_{i}*T*

_{1}, the gradient was analyzed when

*tx*

*reached 3.06 hours, while attempting and adjusting an arbitrary value of*

_{i}*T*

_{1}.

### 3. Training Optimization

*P*, and it was calculated according to Equation (8),

### Results and Discussion

*tx*

*and the work time to reach reference*

_{i}*tx*

*, as a function of an arbitrary*

_{i}*T*

_{1}in the LFCM. The change in both values is shown in Fig. 2. The gradient of

*tx*

*indicates the level of change in*

_{i}*tx*

*for the first 10 minutes after starting the first task, and the working time is the time it takes to satisfy the criteria of skilled workers when performing tasks according to the scenario. The larger the gradient of*

_{i}*tx*

*, the faster the worker’s learning proceeds. As a result of checking the change of*

_{i}*tx*

*for gradient every 10 minutes, the valid range of values was 3.06 hours ≤*

_{i}*T*

_{1}≤6.5 hours. The minimum value of

*T*

_{1}is 3.06 hours, and in the case where 7 hours was exceeded, a reference

*tx*

*of 3.06 hours was not derived. Therefore, the range of possible*

_{i}*T*

_{1}values is considered from 3.06 hours to 6.5 hours. The gradient of

*tx*

*decreases and the work time to reach it increases as*

_{i}*T*

_{1}increases. It can be predicted that, as

*T*

_{1}increases, the time required for a worker to become skilled increases, and the performance becomes saturated in the case of a skilled worker. For the LFCM analysis, 6.5 hours was determined as the reference

*T*

_{1}. The change in time required to work on a unit was analyzed by applying the LCFM to the derived reference

*T*

_{1}.

*tx*

*required for a unit of work.*

_{i}*tx*

*decreased from 4.61 hours to 4.13 hours, and the changes are stagnant during the lunch and break times. Fig. 4 shows the result of analyzing the learning effect over a long period (6 months), as this trend is accumulated.*

_{i}*tx*

*during weekdays were considered. Therefore, a steady decrease in*

_{i}*tx*

*is observed when the worker engages in the work process for a period of 5 days. A sharp increase in*

_{i}*tx*

*appears after the 2-day weekend; however, it does not appear on the graph. As time passes, the gradient of*

_{i}*tx*

*decreases, and the time tends to converge. To better identify these trends, the 5-day values were averaged into a weekly average value, and are shown in Fig. 5.*

_{i}*tx*

*. By fitting this graph, the following equation can be derived.*

_{i}^{2}value of the fitting function was calculated to be 0.99, indicating that the fitting function makes a good representation of the data trend. Considering that the LFCM is in the form of an exponential function, it can be considered acceptable that the derived function converges to 2.98 hours to complete a unit of work as time increases. The fitting function is simpler than the Equations (1)–(7) used for evaluation and can be useful for trend analysis and subsequent training-days optimization.

*x*

*in the LCFM, and the results are listed in Table 2. Because the number of teams is derived as 30, the accumulated number of work units must exceed 30 to fit the total of 900 work units. Table 2 shows that 30 tasks were completed within 16 days. Thus, for exposures below the annual dose limit, 30 teams can complete the dismantling of the radioactive concrete by working in 16 days. If mock-up training using DWS similar to the actual work is performed to gain experience, the work efficiency can be increased while reducing the working time in the radiation work area. Moreover, the efficiency based on the training can be calculated by comparing the accumulated number of work units for 16 days. To analyze the training effect, the increase in the number of work units over training days was analyzed, as shown in Fig. 6. Through this analysis, it can be confirmed how much the workload increases when performing the same task according to the number of training days.*

_{i}*dx*

*/*

_{i}*dtx*

*) between the numbers of the two nearest work units.*

_{i}