**2. State-of-the-Art**

Many researchers in logistics have examined the influence of high-performance logistics practices on organizational performance [1,10,11]. In an attempt to drive performance improvements, managers often struggle with multiple, seemingly conflicting, objectives [1]. Logistics managemen<sup>t</sup> is faced with a tough choice: either strive for efficiency; or strive for effectiveness. Some recent logistics research has suggested that these two performance objectives are mutually exclusive [12]. Performance measures are essential for effective managemen<sup>t</sup> of any organization. Performance measurement provides a needed assessment of service and cost aspects of logistics execution in the supply chain. Specifically, there is little guidance regarding where a specific measure should be used and, more pointedly, where the use of the measure would be less appropriate.

Fugate et al. [13], have presented the model of logistics performance with the concept of simultaneous pursuit of efficiency, effectiveness and differentiation. However, most companies' priorities change over time due to market and competitive dynamics. In light of this business reality, enterprises and managers must be able to identify and select new or different measures consistent with evolving organizational priorities [14].

Muthiah and Huang [15] reviewed and categorized various productivity improvement methods and productivity metrics. For example, Overall Equipment Efficiency (OEE is an established technique in World Class Manufacturing (WCM). It is used as a key performance indicator (KPI) in conjunction with lean manufacturing efforts to provide a quantifiable measurement of success. There are a few examples of the performance evaluation of manufacturing systems with the use of the OEE metric, including [16,17], but without considering the efficiency of the transport subsystem.

Muñoz-Villamizar et al. [18], have used OEE to evaluate the effectiveness of urban freight transportation systems and a framework for Overall Transport Efficiency (OTE) based on OEE factors was proposed by Dalmolen et al. [19]. McCalion [20] ask the question: is OEE relevant to logistics managemen<sup>t</sup> and Automated Guided Vehicle (AGV) operations?

Hayes [21] suggests that the OEE can be used for eliminating the ripple effect caused by stopped vehicles and along with Six Sigma [22] can be used as a measure for World Class Logistic (WCL).

Comparing WCL with WCM, they have a lot in common. The common area is related to intra-logistic in manufacturing systems. Intra-logistic performance has a grea<sup>t</sup> influence over the manufacturing performance, because of inter-operational breaks which have a grea<sup>t</sup> impact on materials flow in flexible manufacturing system [23].

The literature review only shows a few publications on the design methodology of AGV systems and most of the them use simple KPIs as metrics [24–29]. At the time of preparing this paper, no publication was found concerning the detailed assessment of the impact of the AGVs system on the manufacturing process e ffectiveness that includes the OEE metric. Therefore, the studies have been undertaken in order to elaborate this problem in terms of transport and production e ffectiveness and to strengthen the logistics potential of the organization.

#### *2.1. Issues Related to FMS and AGV*

The flexible manufacturing system (FMS) is a fully automated production system that interconnects machines and workstations with the logistics equipment, where the entire manufacturing process is coordinated by the digital control systems such as Computer Numerical Control (CNC) or Programmable Logic Controller (PLC). Such flexible, automated manufacturing systems are intended for tasks of large typological diversity, high complexity, ensuring on-time delivery and minimal manufacturing costs, while production is unpredictable, being organized in small batches, with frequent changes [30].

The FMS has been studied over the last couple of decades and the researchers have found a variety of problems, which can be distributed in three major categories: workshop design, transportation network design and scheduling problems [31,32]. Di fferent methods were used to solve them, including mathematical (linear, constraints, stochastic) programming, combinatorial optimization, Petri nets and scenario analysis, but computer simulation, especially Discrete Event Simulation (DES), is the most universal and widely used one [33], e.g., for the design of manufacturing systems [27], e fficiency [9] and stability analysis [34] of production systems and the design of warehouse transportation systems with Automated Guided Vehicles (AGVs) [35–37].

The AGVs are classified as service robots for professional purposes in manufacturing environments and broad review of AGV is presented in [6,28].

Modern AGV vehicles are characterized by precision of operation, speed of movement and high reliability. They can have various equipment to perform numerous transport tasks, such as transporting pallets and containers, pulling trailers with cargo, lifting with a forklift or manipulating details using an integrated robot arm. Examples of AGV carriages are shown in Figure 1.

(**a**) (**b**) (**c**)

**Figure 1.** Examples of Automated Guided Vehicles (AGVs), (**a**) pulling trailers with cargo, (**b**) transporting pallets, (**c**) lifting loads with a forklift [38].

In comparison to the other solutions of transport systems, AGV vehicles show many advantages including [5,6]:

• they do not require an operator's service, which allows reducing the labour costs,


Typical features of AGV are related to the following parameters, as [5]:


AGV uses electric drive and efficient batteries, however, it requires periodic recharging. Depending on the battery capacity and load, a typical work cycle includes 8–16 h of work and 4–8 h of battery charging, which takes place completely automatically. Some solutions for recharging the battery during short interruptions in the work of AGV (Opportunity Battery Charging) can be found. There are also solutions based on manual or automatic replacement of a depleted battery with a new fully charged battery (Battery Swap). This action takes about 10 min and allows to take full advantage of AGV's working time but requires a more advanced service system and additional battery packs.

The design of a transport system based on AGVs requires an advanced navigation system and appropriate delineation of transport routes and reloading points. Based on the technique, various navigation systems are used, such as [37]:


Various methods can be used to design AGV systems, including mathematical programming methods, heuristic methods, Petri nets and computer simulations. These methods are used in order to improve the transport network in terms of criteria, for minimizing the length of transport routes, maximizing the production flow, scheduling transport tasks, number and location of transhipment points, parking zones and others [28,39].

Transport routes can be one- or two-way. Due to the reduction of the risk of collision, one-way roads in the form of closed loops, which enable cyclical transport operations, are the most commonly used [22]. In this case, it is easier to develop traffic control algorithms than in the case of two-way traffic, which requires additional passing and parking zones. Therefore, during the design of the AGV system, the most frequently used zones are defined including specific segments of the route (Segmented Flow Configuration) and individual transport loops (Single Loops) in a given segment. The advantages of such a solution are related to [40]:


In turn, some drawbacks are connected with [35]:


When designing the AGV transport system, the most important problem is determining the number of vehicles needed to achieve the required production volume or the minimum number of vehicles required to obtain the optimal production volume [36,40].

There are a lot of situations in which the AGV system may stall because of a deadlock. A variety of deadlock-detecting algorithms are available in literature [41], but these methods work mainly for manufacturing system where the network layout is simple and uses only a small number of AGVs. The paper [42] discusses the development of an e fficient strategy for predicting and avoiding the deadlocks in a large scale AGV systems. The integration of the scheduling of production and transport tasks tends to also be problematic because of computational complexity [43,44]. In initial papers, the transportation times between machines have not been considered. Their authors claimed that because transportation times are very small in comparison with the processing times, they are negligible [45]. On the other hand, in recent decades, the more researchers have been attracted by some issues that the transportation times were considerable and ignoring them can have impacts on the solution of scheduling problems.

#### *2.2. Evaluation of FMS and AGV*

The performance of the AGV logistics system can have a grea<sup>t</sup> influence over the performance of the whole FMS system; therefore, a performance evaluation should be conducted. The key performance indicators (KPI) of the production system include [16,46]:


Work e fficiency and the use of the means of production can be expressed by using the OEE metric that depends on three factors: availability, performance and quality [16].

$$\text{OEE} = \left( \text{Availability} \right) \times \left( \text{Performance} \right) \times \left( \text{Quality} \right) \tag{1}$$

Availability is the ratio of the time spent on the realization of a task to the scheduled time. Availability is reduced by disruptions at work and machine failures.

$$\text{Availability} = \frac{\text{available work time} - \text{failure time}}{\text{scheduled time}} \tag{2}$$

Machinery failures may cause severe disturbances in production processes and the loss of availability. Inherent availability can be calculated with Formula (3).

$$\text{Availability} = \frac{\text{MTBF}}{\text{MTBF} + \text{MTTR}} \tag{3}$$

where:


The OEE metric was developed for single component maintenance. In the case of complex systems including serial or parallel subsystems the availability is changed. For the series system to be available, each subsystem should be available. For the parallel system to be available, whichever subsystem should be available.

Performance is the ratio of the time to complete a task under ideal conditions compared to the realization in real conditions; or the ratio of the products obtained in reality, to the number of possible products to obtain under ideal conditions. Performance is reduced (loss of working speed) by the occurrence of any disturbances, e.g., human errors.

$$\text{Performance} = \frac{\text{ideal cycle time}}{\text{real cycle time}} = \frac{\text{real number of products}}{\text{ideal number of products}} \tag{4}$$

Quality is expressed by the ratio of the number of good products and the total number of products.

$$\text{Quality} = \frac{\text{number of good quality products}}{\text{total number of products}} \tag{5}$$

To compare the influence of the AGV logistic system over the manufacturing system, we will consider different OEE factors. However, according to the lean manufacturing paradigm, the flow of production through bottlenecks is the most important, therefore some equipment should be fully utilized whilst other equipment does not require full utilization. The literature review [16,46] indicates that OEE metrics are lacking at complex manufacturing systems and the factory level. In order to address this gap, an overall throughput effectiveness metric can be used [47]. It measures the factory-level performance and can also be used for performing factory-level diagnostics such as bottleneck detection and identifying hidden capacity. It also accounts for subsystems processing multiple products. Any factory layout can be modelled using a combination of the predefined subsystems (serial, parallel), which allows a determination of the Overall Factory Effectiveness (OFE). Note that the OEE equation can be further simplified as [46,47]:

$$\text{OEE} = \frac{\text{Actual throughput (units) from equipment in total time}}{\text{Theoretical throughput (units) from equipment in total time}} \tag{6}$$

By extending this definition to the factory level, we have Overall Factory Effectiveness (OFE):

$$\text{OFE} = \frac{\text{Actual throughput (units) from factory in total time}}{\text{Theoretical throughput (units) from factory in total time}} \tag{7}$$

Similarly, the Overall Transport Effectiveness (OTE) can be defined:

$$\text{OTE} = \frac{\text{Actual throughput (units) from vehicle in total time}}{\text{Theoretical throughput (units) from vehicle in total time}} \tag{8}$$

#### **3. Description of the Problem—Materials and Methods**

Let us consider a production system with eight automated machine tools, such as a CNC machining centre, which performs a two-stage process of machining a family of typical machine parts, like sleeves or discs of different sizes.

The machines are arranged as shown in Figure 2, which allows the series-parallel flow of production.

**Figure 2.** The schema of Flexible Manufacturing System (FMS) with AGV transport.

The system also includes an Automatic Washing Station and Inspection Station as well as a Storage System with an automatic rack stacker. Randomly generated various production orders are delivered to the system, di ffering in the duration of the operation (from 2.5 to 15.2 min). As a means of product transportation devices, several AGV vehicles are used, which will move along the planned transport routes. We assume that the manufacturing process meets the lean manufacturing, i.e., the flow of a single product and minimal bu ffers capacity to limit production in progress.

When designing a production system, we strive to achieve maximum production e fficiency and, in particular, maximum utilization of machines and devices constituting bottlenecks in the manufacturing process. Other machines and devices will usually be used to a lesser extent, but they are necessary for the production process. By introducing changes to the model, we can analyse the formation of bottlenecks in the production system and their impact on the production volume. This allows, i.e., to determine the required storage capacity and capacity of the transport system. Particularly, the most interested issue is the impact of the number of AGV transport resources on the production e fficiency of the entire system. For this purpose, a simulation model was developed in the FlexSim 2018 environment, shown in Figure 3.

**Figure 3.** Simulation model of the FMS with AGV transport.

Initially, the reference system consisted only of machines, without taking transport into account. It represents the manufacturing system in ideal conditions. At the next step, transport-related constraints were added to the model. The layout takes into account the dimensions of individual model objects and the distance between them. According to the recommendations given in the literature, the model uses unidirectional transport routes forming three main loops. Several control points have been introduced to designate places of loading and unloading as well as parking spaces. For the most used intersections, control zones were used to reduce the risk of collisions and blockages.

Typical parameters of AGV were assumed, including the speed of 2 m/s and a loading/unloading time of 30 s. A FIFO (First In First Out) control strategy was applied. In addition, the warehouse system was expanded by adding components such as the high storage warehouse with an automated storage retrieval system (ASRS).

#### **4. Results of the Simulation Experiments**

The developed model was used to conduct a series of simulation experiments. In subsequent experiments, the number of AGV vehicles from 0 to 8 was changed. The simulation time of 24 h was assumed as the time of automatic maintenance of the entire system. A random generation of production orders was assumed according to the exponential distribution with the expected value of 100 s. As a result of a lack of data and the need for simplification, the retooling of the system, charging of AGV batteries and the failure of machines and vehicles was omitted. As part of each experiment, 30 simulation runs were carried out. The results are presented in Table 1 and Figure 4, respectively. Due to the stochastic parameters of the model, the production value Pavg obtained in the experiment is a random variable with a distribution close to normal.

**Figure 4.** The relationship between average production Pavg and the number of AGV used for 24 h of simulation.

The first row in the Table 1, where the number of AGV amount is 0, represents the reference system consisting only of machines and not taking transport into account (transport time equal to zero).

Figure 4 presents the relationship between average production Pavg and the number of AGVs used for transport in the form of a box-and-whisker plot.

The box represents a 95% confidence interval, which means the average production volume is within this range with a probability of 95%, whilst, whiskers represent the maximum and minimum value of production obtained in a given experiment (there is a small spread of results, therefore some of the boxes in the chart are very small).

In the Figure 4, it can be seen that initially the increase of the AGV number results in a rapid increase of obtained production volume. On the other hand, increasing the AGVs number above 5 units results in a slight increase in production, as more vehicles are used to a lesser extent. A similar phenomenon is described in the literature [36] as the e ffect of the mutual blocking of AGVs.


**Table 1.** The results of simulation experiments (Average production completed Pavg, in [pieces] for 24 h of work, 30 simulation runs in each experiment, 95% confidence level).

To eliminate them, it would be necessary to use multi-lane transport routes or passages. The excessive increase in the number of AGVs is in turn associated with high costs and brings little e ffects and a relatively small increase in production e fficiency [34] with a drop in the e ffectiveness of AGV being used.

The comparison with the results obtained in the ideal conditions (AGV = 0, Pavg = 561 pc.) shows a grea<sup>t</sup> di fference with the other results. For example, for 5 AGVs system, there was a Pavg = 441.5 pc., average machine utilization of about 70% and average AGVs utilization of about 50% (from the range of 28–71% for AGV1 and AGV5). The increase of the AGV number caused very little increase in machine utilization and a considerable decrease in the utilization of the additional AGVs.

That problem requires a more detailed investigation, but the traditional metrics for measuring productivity as throughput or utilization rate are not very helpful for identifying the problems and underlying improvements needed to increase productivity. In this situation, a more rigorously defined productivity metric is needed [44]. Therefore, in this case, OEE metrics can be used, which take into account equipment availability, breakdowns, performance (reduced speed, idling) and quality (good and bad quality products).
