CC BY
141Russian Journal of Logistics and Transport Management, Vol.2, No.1, 2015
©Yulia Panova1 and Olli-Pekka Hilmola2
Petersburg State Transport University 2Lappeenranta University of Technology, Kouvola Unit
DYNAMIC BOTTLENECKS IN HANDLING AND STORAGE SYSTEMS
Abstract
The development of industrial engineering and production systems are highly appreciated under the demand of Russian customers in the current economic and political situation, e.g. deprivation from several import markets. In these circumstances, issues related to the formation of process systems are gaining their importance. The article considers the objective of reaching the smooth and continuous material flow in the handling and storage system of the plant, as well as the problems of bottlenecks optimization.
Keywords: bottlenecks, simulation, handling, and storage system.
1 Introduction
The analysis of the different industrial engineering and technological systems shows that the large part of logistics costs of goods’ production are originating from operations of warehousing and transportation (Rantasila, 2013). Broadly speaking, within the overall time allocated for storage, shipping, and manufacturing operations, the time, which is required to manufacture the product, is on average 2-5%. Meanwhile, over 95% of the time turnover accounts for logistics operations related to supplying and distribution (Balalaev et al., 2008). That is the management of material flows.
Similarly, in the overall time, which goods spend under the technological cycle, the large part is connected with the transportation and waiting times. The reduction of these periods may accelerate the capital turnover, respectively, increasing the profits obtained per unit of time, and reducing the final costs of the product.
In light of the current conditions, the issues related to industrial engendering and production plants in Russia are gaining their importance. The timeliness of the questions can be partly explained by the Russian sanctions on imports from Australia, Europe and North America, and, as a result, the necessity for the development of the national production economy.
A particular attention can be paid to the assembling production or process plants that can be characterized by several peculiarities such as the need for synchronization by number, pace and time of deliveries of several components. The problems of managing the lean operation, bottlenecks identification and improvements of the systems are well described in the work of Goldratt et al.
11
(2004). The book has become the standard for management and optimization of bottlenecks (Jacobs, 1984) as its initial edition was published in the 1980’s (see journal article by this lead author: Goldratt, 1988). Discipline of theory of constraints has developed during the decades to the wide spectrum of application areas such as supply chain management (Tsou, 2013), accounting (Hilmola and Gupta, 2015; Myrelid and Olhager, 2015), strategy (Costas et al., 2015) and project management (Mabin and Balderstone, 2003). These of course together with production scheduling and control.
Dynamic or moving bottlenecks are perceived as one of the difficult to identify in process plants (King, 2011). The reason for that is that the bottleneck can move in the process cycles (typically named as wandering bottleneck). Due to this fact, the analysis of qualitative and quantitative cross-sectional data, as well as stocks in the technological process is reasonable with the use of simulation modeling.
A computer simulation allows to practice the mental model and to explain the behavior implicitly in the structure, by which the model is constructed. Simulations of different systems are famous nowadays. Examples would be wind tunnel tests of aircraft design, weather patterns, and the depletion of oil reserves. Economists and social scientists also have used simulation to understand, how energy prices affect the economy. The developed models with sufficient accuracy describe the systems under study, allowing to simulate the behavior of those objects, especially when the experiments with them are impossible or dangerous. Thus, the simulation models are created to answer the questions of type ‘what if ...’, i.e. to investigate possible scenarios for the development of systems (Karpov, 2005).
In the article, the simulation is used to analyze the handling and storage system. The statistics collected from the model served as the basis for its rational arrangement that, in turn, was the primary prerequisite for the efficient functioning of the plant.
2 Formalization of the task
Organization and management of material flows are provided by the handling and storage system (HSS). The system, in turn, serves two primary functions: the process of flows and their storage. The activities that happen within the system were described through the simulation modeling. As the simulation environment, the modern software AnyLogic was applied. The first version of the program was released in 2000. At the root of its creation is a group of scientists from St. Petersburg State Polytechnical University. Their success of research initiated the establishment of the Russian company XJ Technologies with headquarters in St. Petersburg. Thanks to the network of 27 distributors, which covers six continents from Australia to North America, program Anylogic employs more than 15,000 users in 60 countries (XJ Technologies, 2012). The
12
reason for the high demand for this simulation environment is due to the following reasons. The program allows to create models based on any of the paradigms of simulation: system dynamics, discrete event, agent-based modeling, as well as their combination within a single model (XJ Technologies, 2012).
It should be noted that the approaches mentioned before imply different levels of abstraction and detailing of the processes. For example, agent-based modeling has the highest level of abstraction, while discrete event simulation details all processes. As a rule, the choice of the approach depends on the goals of researchers. The goal for modeling HSS was to analyze the arrival rate of entities (materials) and rhythms of work of different steps of the system. In addition, to consider the utilization of service blocks and aggregates. What is more important is to identify bottlenecks, critical levels of buffers so as to make the recommendations for the improvement of the throughput of the system.
For the description of complex processes, such as the working process of HSS, the most suitable way of modeling was chosen. That is, the discrete-event simulation (DES), which, according to studies performed, was the most prevalent in the field of production and business (Jahangirian et. al., 2010). This simulation method was developed by Geoffrey Gordon in 1960. By nowadays, the method has received a broad scope of applications: From logistics and queuing systems to transport and production systems (Martinez et al., 2001; Moon and Phatak, 2005; Goti et al., 2011). Depending on the goal and chosen simulation approach (discrete event), the parameters to describe the system was outlined. For the simplicity of calculations, the system was represented by three steps (Figure 1).
Inflows
Step A Step B hfM Step C
Outflows
Buffer
Buffer
Fig. 1. The scheme of the HSS.
The number of the main parameters used for the solvent of tasks is as follows: 1) Arrival rate of the entities, 2) The duration of their processes on each step, and 3) The rule of handling entities.
3 Simulation modeling of the handling and storage system
The DES model was designed with AnyLogic 6.9 in the form of blocks that process the entities, following the specified parameters and connections between them. The connections between blocks determine the sequence of operations. The blocks of the discrete-event model are represented by the objects of Enterprise Library (Figure 2).
13
Fig. 2. Process model of queuing system developed in AnyLogic 6.9.
This model contains following blocks:
Object of the Enterprise Library Functions
Source Generates an entity in accordance with the law of the random variable.
Service Captures the specified number of requests for resources, delaying the entity, and then releases resources that were caught by an entity. The object is an equivalent to a sequence of objects, such as Seize, Delay, and Release.
ResourcePool Defines a set of resources that can capture and released by entities through used objects (e.g. Seize, Release, and Service).
Sink Removes the entities from the system (end-point of the entities flow).
TimeMeasureStart TimeMeasureEnd TimeMeasureEnd with TimeMeasureStart is a pair of objects that allow to measure the time spent by entities between two points of the process diagram. Usually with the help of these objects, the time that entities spend in the system or some sub-process is measured.
The performance of entities processing was characterized by a triangular distribution. Time of entities arrival at the system was exponentially distributed. The rule of entities service is FIFO (First-In-First-Out).
The service time of entities in the three blocks of Service was defined by a triangular distribution law (e.g. triangular( 0.05, 0.1, 0.2)*minute(), triangular ( 0.4, 0.5, 0.6)*minute(), and triangular( 0.3, 0.4, 0.5)*minute(), respectively). For example, triangular( 0.05, 0.1, 0.2 )*minute() means that that minimum is 0.05 min., maximum is 0.2 min., and the most likely is 0.1 min.
Entities arrive at the system with the rate 2/minute().
The duration of computer simulation prescribed for 100 days with 8 hours shifts (48 000 minutes).
The efficiency of the queuing system was characterized by primary parameters:
- Coefficient of loading of resources;
- Average number of entities in the queue;
- Number of entities served by the system (throughput);
- Average time that entity spent in the queue (found for the bottleneck);
- Average time that entities spent on the several steps;
- Average time that entities spent in the system.
14
The parameters mentioned above allowed to create recommendations for the handling and storage system as a whole, as well as for the several subprocesses.
For example, the experiment of the model shows the bottleneck that is depicted by the coefficient of loading of resources (Figure 3).
Fig. 3. Coefficients of resources utilization.
As can be seen from the Figure 3, the bottleneck in the system is the second step (Service1) with the high loading coefficient of ResourcePool 1 (0.998). Due to the slow work of this block, there is a queue of entities in front of Service1 line. The location of entities in queues on this step or others signalizes about the need for additional storage capacity in the manufactory or necessity to increase the efficiency of the service lines.
In case of the first option, the capacity of the stock should be equaled to the maximum length of the queue. However, in most of the time, the capacity of the storage will not be fully utilized (during the year time). If there is a chance to allow the stocks to be overloaded in the several pick periods, then the capacity of the storage is possible to reduce. Based on the experiments with the model, the following statistics have been collected (Figure 4).
Fig. 4. Statistics collected from the model.
15
All statistics have been summarized in Table 1.
Table 1
Parameters of the model.
Description of the parameter Number of the ste ps
1 2 3
Coefficient of resources utilization 0.233 0.998 0.798
Average number of entities in the queue 0.769 72.664 0.22
Average time that entity spent in the queue, min 36.27
Average time that entities spent on the several steps, min 0.15 36 0.4
Number of entities served by the system 95 793
Average time that entities spent in the system, min 37
The use of graphical statistics allows making decisions on the capacity of the storage, depending on the number of days, when it will be overloaded. For example, the required storage capacity of Step 2 (Service1), would be 200 units in the following case. Space for the stock will be fully loaded only one day, while the other 99 days the storage capacity will be underused (the coefficient of resources utilization/loading (L) would be 0.01). If to consider L=0.08, then the required storage capacity is 160 units that allow to eliminate the irregular pace of arrival of entities and production department (e.g. Service1; Figure 4).
The coefficient of loading resources at Step 2 (Resource Pool 1) is high (0.997). It means the assets are employed at 0.997 at the service time of 0.5 minutes (on average). Resources are very busy; therefore, any spare capacity is not available. In other words, any equipment failure will lead to the fact that the storage will cease to cope with the flow of entities. The queue will increase dramatically, and, thus due to long delay times at Step 2, the following technologic processes at Step 3 will be suspended because of the lack of items/entities. The increase of service time at Step 2 (Service 1) by 0.1 minutes to 0.6 minutes on average will lead to the reduction of the throughput of the system. It will be reduced by 18% (from 95783 to 78344 entities). The dependence of the service time at the bottleneck (Step 2) and the length of the queue in front of the service line are represented in Table 2.
Table 2
The dependence of the service time and size of the queue.
Average number of entities in the queue, min Average service time in Service 1, min Average time that entity spent in the queue, min
7947 0.6 4058
72.664 0.5 36.27
2.121 0.4 1.18
0.923 0.3 0.512
0.586 0.2 0.5
16
As can be seen from Table 2, both parameters (e.g. the length of the queue and the average time that entity spent in the queue) increase exponentially. On the grounds of dependencies from the collected statistics, one can conclude that the only way to improve the performance of the system is to reduce the duration of the operations at Step 2 (Service 1). The lessening of service time at the bottleneck (Service 1) can be achieved by the following option: That is the reduction of the service time per unit, in other words the processing of three units/entities instead of two at the same time (i.e. the number of resources/equipment can be increased to three, ResourcePool 1).By doing so, the service rate increases, reducing the number of entities in the queue to maximum six entities or average 1.515 (Figure 4, Table 3).
Table 3
Parameters of model with the increased number of Resource Pool 1.
Parameters of the model Number of the steps
1 2 3
Coefficient of resources utilization 0.233 0.333 0.799
Average number of entities in the queue 0.777 1.515 2.815
Average time that entity spent in the queue, min 0.509
Average time that entities spent on the several steps, min 0.15 0.5 1
Number of entities served by the system 95 887
Average time that entities spent in the system, min 1.8
Fig.5. The statistics collected from the model, including improved parameters of Step 2.
17
At the beginning of the analysis of the efficiently of the process of entities, Step 2 was the bottleneck of the whole handling and storage system. The reduction of service time at this Step resulted in the ‘transition’ of the bottleneck to the next block (Step 3). The right management of the new bottleneck at Step 3 also can lead to a better performance of the system. Sometimes the reason for low output ratio can relate to the limited buffer size before bottleneck that constraints utilization of bottleneck resource and system as a whole. Thus, the introduction of additional storage space at Step 3 positively influenced the production of finished goods (e.g., the throughput of the whole system). At the same time, the size of inventory holding space should be considered properly, since its usefulness is high at small volumes, while with a larger amount of work-in-process, it decreases (Lambrecht et al., 2012).
4 Conclusions
The provided study of the handling and storage system of the plant showed that the principal instrument for the research and generation of the recommendations can be the method of simulation modeling. Analytical modeling is appropriate to the analysis of the sub-process, while simulation allows understanding the interdependencies between the subsystems and finding the characteristics of the whole system.
Based on the queuing theory the analytical calculations and statistics were analyzed. The experiments with the model identified the bottleneck that was further treated with increasing capacity (lower processing cycle-time). One of the reasons for the inefficient parameters of the system was the slow service time at Step 2 of the plant. It resulted in the long queues and delays/lacks of units in the further steps of the process. Resources were very busy (with the loading coefficient of 0.997). It meant that any equipment failure would lead to the fact that the storage would cease to cope with the flow of entities (e.g. the queue will increase dramatically). On the grounds of the found dependencies from the collected statistics, it was concluded that the only way to improve the performance of the system is to reduce the duration of the operations at Step 2. In the experiment was clearly shown that cycle-time improvement of constraint resource reduces queue (inventory) amount in front of resource, and this seems to decline substantially after some threshold is reached in constraint performance.
The undertaking to improve the efficiency of Step 2 (e.g., the increase of resources/handling equipment) resulted in the shift of the bottleneck from the current Step 2 to Step 3. So, as to manage the bottleneck at the Step 3, the organization of the entities processing can be improved by the introduction of additional storage space. By doing so, the parameters of the system significantly improved, leading to the growth of the throughput of the plant. In conclusion, it should be noted that the received estimates are approximate. Therefore,
18
corrections in the simulation model for the case of the real production or industrial engineering system would be required.
References
Balalaev, A.S., Kochemasova, A.V., & Tretiak, S.N. (2008). Transport and Warehouse Logistics Ensuring. Khabarovsk: Publ FESTU, Russia, 140.
Costas, J., Ponte, B., de la Fuente, D., Pino, R. & Puche, J. (2015). Applying Goldratt’s theory of constraints to reduce the bullwhip effect through agent-based modeling. Expert Systems with Applications, 42(2), 2049-2060.
Goldratt, E. M. (1988). Computerized shop floor scheduling. International Journal of Production Research, 26(3), 443-455.
Goldratt, E. M., Cox, J. & Whitford, D. (2004). The Goal: a Process of Ongoing Improvement. MA: North River Press, USA, 390.
Goti, A., Zabaleta, N., Garcia, A., Ortega, M., & Uradnicek, J. (2011). Optimize reordering of critical raw materials and parts. Hydrocarbon Processing, 90(5), 79-81.
Hilmola, O-P. & Gupta, M. (2015). Throughput accounting and performance of a
manufacturing company under stochastic demand and scrap rates. Expert Systems with Applications, 42:22, 8423-8431.
Jacobs, R. F. (1984). OPT uncovered: many production planning and scheduling concepts can be applied with or without the software. Industrial Engineer, 16(10), 32-41.
Jahangirian, M., Eldabi, T., Naseer, A., Stergioulas, L. K., & Young, T. (2010). Simulation in manufacturing and business: A review. European Journal of Operational Research, 203(1), 1-13.
Karpov, J.G. (2005). Simulation Modeling Systems. Introduction to Modeling with AnyLogic 5. St. Petersburg, Russia, 400.
King, P. L. (2011). The bottleneck conundrum. Industrial Engineer. 43(1), 41-46.
Lambrecht, M., Creemers, S., Boute, R., & Leus, R. (2012). Extending the production dice game. International Journal of Operations & Production Management, 32(12), 14601472.
Mabin, V. J. & Balderstone, S. J. (2003). The performance of the theory of constraints methodology: Analysis and discussion of successful TOC applications. International Journal of Operations and Production Management, 23(6), 568-595.
Martinez, J. C., Trani, A. A., & Ioannou, P. G. (2001). Modeling Airside Airport Operations Using General-Purpose, Activity-Based, Discrete-Event Simulation Tools. Transportation Research Record: Journal of the Transportation Research Board. 1744, 65-71.
Moon, Y.B. & Phatak, D. (2005). Enhancing ERP system's functionality with discrete event simulation, Industrial Management & Data Systems, 105(9), 1206-1224.
Myrelid, A. & Olhager, J. (2015). Applying modern accounting techniques in complex manufacturing. Industrial Management and Data Systems, 115(3), 402-418.
Rantasila, K. (2013). Measuring Logistics Costs -Designing a Generic Model for Assessing Macro Logistics Costs in a Global Context with Empirical Evidence from the Manufacturing and Trading Industries. Publication Series A-8/2013. Dr. Dissertation, University of Turku, Turku, Finland.
Tsou, C.-M. (2013). On the strategy of supply chain collaboration based on dynamic inventory target level management: A theory of constraint perspective. Applied Mathematical Modelling, 37(7), pp. 5204-5214.
XJ Technologies (2012). About the company http://www.xjtek.ru Accessed 02.04. 2015.
19
