To increase the flexibility of application of continuous conveyor systems, a completely decentralized control system for a modular conveyor system is introduced in the paper. This system is able to carry conveyor units without any centralized infrastructure. Based on existing methods of decentralized data transfer in IT networks, single modules operate autonomously and, after being positioned into the required topology, independently connect together to become a functioning conveyor system. Parallel to the development of the decentralized control system, identical square modules were developed, which in a compact unit contain all of the features necessary to function as a switch, junction or linear conveyor section. To fulfill this task, every module is equipped with an RFID identification system, sensors, a multi-directional drive, and a microcontroller-based control unit that executes the control algorithm.
The following functions can be performed by these modules with the help of the innovative control algorithm:
• Independent generation of the topological map in the form of routing tables
• Recognition of an incoming conveyor unit and identification of the destination address
• Planning of the path to the destination taking into consideration conveyor units already located in the system
• Protection against collisions and deadlocks, and transportation of the conveyor unit to the next module
• Autonomous regulation of the injection rate to ensure the highest possible throughput
The throughput performance of the control algorithm developed here was analyzed by simulating representative topologies. Furthermore, it was proven that under certain conditions, despite the conveyor routes being used in multiple directions, a situation can never arise where conveyor units block each other and the flow of material comes to a halt in the form of a deadlock.
Contents
Kurzfassung
Abstract
1 Introduction
1.1 Motivation
1.2 Aim of the dissertation
1.3 Structure of the dissertation
2 Centralized material flow controls
2.1 Classification of material flow controls
2.1.1 Control concepts
2.1.2 Structure of the control systems
2.1.3 Signal processing
2.2 Tasks of the control levels in the material flow automation
2.3 Operating method of conventional PLC controls
2.3.1 Field of application of a PLC
2.3.2 Structure of a PLC
2.3.3 Programming languages
2.3.4 Operating concepts
2.3.5 PLC networks
2.3.6 Limits of and alternatives to centralized controls for material flow systems
3 Decentralized control systems
3.1 Definition of the term ”decentralized” in material flow
3.2 Research projects in decentralized material flow controls .
3.2.1 The Internet of Things
3.2.2 MATVAR
3.2.3 Transport system in analogy to routing in data networks
3.2.4 Need for research to achieve complete decentralization
3.3 Decentralized control of IT networks
3.3.1 The OSI reference model
3.3.2 LAN technology
3.3.3 Transport protocol
3.3.4 Routing in networks
3.3.5 Decentrally controlled information vs. material flows
4 Completely decentralized autonomic continuous conveyor system
4.1 Overview and general assumptions
4.1.1 Requirements for a completely decentralized system
4.1.2 Determination of the physical features
4.1.3 Application example
4.2 Control concept
4.2.1 Decentralized generation of topological information
4.2.2 Identification of the conveyor unit
4.2.3 Routing and route reservation
4.2.4 Transportation of conveyor units
4.2.5 Deadlock avoidance
4.3 Throughput analysis
4.3.1 Simulation environment
4.3.2 Throughput calculation
4.3.3 Topology analysis
4.4 Throughput regulation
4.5 Interfaces to the environment
5 Technical implementation
5.1 Introduction of the ”Flexconveyor”
5.2 Construction
5.2.1 Base plate with lifting mechanism
5.2.2 Diverter with integrated RFID antenna
5.2.3 Roller arrangement and sensor system
5.3 Control of the Flexconveyor
5.3.1 Electrical connection
5.3.2 Control procedure
5.4 Connection of several modules to the topology
6 Summary
References
List of figures
Kurzfassung
Stephan Mayer
Entwicklung einer vollständig dezentralen Steuerung für modulare Stetigförderer mit veränderbaren Topologien
Um die Flexibilität und den Einsatzbereich von Stetigförderanlagen zu erhöhen, wird in der vorliegenden Arbeit eine vollständig dezentrale Steuerung für ein modulares Stetigfördersystem vorgestellt, welches Fördereinheiten wie beispielsweise Kleinladungsträger ohne jegliche zentrale Infrastruktur befördern kann. Basierend auf existierenden Methoden zum dezentralen Da-tentransport in IT-Netzwerken agieren die einzelnen Module autonom und koppeln sich nach der Positionierung zur benötigten Topologie selbständig zum funktionierenden Fördersystem zusammen.
illustration not visible in this excerpt
Figure 0.1: ¨Ubertragung der Methoden aus IT-Netzwerken für ein dezen-trales, modulares Stetigfördersystem
Parallel zur Entwicklung der dezentralen Steuerung wurden baugleiche, quadratische Module entwickelt, welche als kompakte Einheit sämtliche Funktionen besitzen, um als Verzweigung, Zusammenführung oder einfache Förderstrecke zu fungieren. Dafür wird jedes Modul mit einem RFID-Identifikationssystem, Sensoren für die Positionserkennung der Fördereinheiten, einem Förderantrieb zum Transport in vier horizontale Bewe-gungsrichtungen und einer Recheneinheit, welche den Steuerungsalgorithmus ausführt, ausgestattet.
Folgende Funktionen können die Module mit Hilfe des neuartigen Steuerungsal-gorithmus ausführen:
- Selbständige Erzeugung der Topologielandkarte in Form von Routingta-bellen
- Erkennung einer ankommenden Fördereinheit und Identifikation der Zieladresse
- Planung der Route bis zum Ziel unter Berücksichtigung der bereits im System befindlichen Fördereinheiten
- Absicherung gegen Kollisionen und Deadlocks und Transport der Fördereinheit zum nächsten Modul
- Selbständige Regulierung der Einlastung von Fördereinheiten, für einen höchst möglichen Durchsatz
Der entwickelte Steuerungsalgorithmus wurde in einer Simulation für repr¨asen-tative Topologien auf seine Durchsatzleistung untersucht. Weiterhin wurde ein Nachweis erbracht, dass unter bestimmten Bedingungen trotz Nutzung der Förderstrecken in mehrere Richtungen, niemals eine Situation entstehen kann, in der sich Fördereinheiten gegenseitig blockieren und es zum Stillstand des Materialflusses in Form eines Deadlocks kommt.
Abstract
Stephan Mayer
Development of a completely decentralized control system for modular continuous conveyors
To increase the flexibility and range of application of continuous conveyor systems, a completely decentralized control system for a modular conveyor system is introduced in the following dissertation. This system is able to carry conveyor units (for example, small load bearers) without any centralized infrastructure. Based on existing methods of decentralized data transfer in IT networks, single modules operate autonomously and, after being positioned into the required topology, independently connect together to become a functioning conveyor system.
Parallel to the development of the decentralized control system, identical square modules were developed, which in a compact unit contain all of the features necessary to function as a switch, junction or linear conveyor section. To fulfill this task, every module is equipped with an RFID (radio frequency identification) identification system, sensors for the position detection of conveyor units, a multi-directional drive to transport conveyor units in four horizontal directions, and a microcontroller-based control unit that
illustration not visible in this excerpt
Figure 0.2: Decentralized methods of IT networks as basis for a decentralized, modular conveyor system executes the control algorithm.
The following functions can be performed by these modules with the help of the innovative control algorithm:
- Independent generation of the topological map in the form of routing tables
- Recognition of an incoming conveyor unit and identification of the destination address
- Planning of the route to the destination taking into consideration conveyor units already located in the system
- Protection against collisions and deadlocks, and transportation of the conveyor unit to the next module
- Autonomous regulation of the injection rate to ensure the highest possible throughput
The throughput performance of the control algorithm developed here was analyzed by simulating representative topologies. Furthermore, it was proven that under certain conditions, despite the conveyor routes being used in multiple directions, a situation can never arise where conveyor units block each other and the flow of material comes to a halt in the form of a deadlock.
1 Introduction
1.1 Motivation
The internet and globalization lead to the fact that not only large corporations but also small and medium-sized corporations have to compete in the international market. The resulting changes in production have a direct effect on the processes of the in-plant flow of products and materials, and new demands on intralogistics emerge.
The biggest changes in the internal material flow are indicated by shorter project durations and declining batch sizes per order. E-commerce plays an enormous role by offering the possibility of a large variety of products in the shortest amount of time. Therefore, each company has to release new products to the market at an increasing rate in order to set itself apart from competing products. The simplification of the ordering process to a click has resulted in, among other things, a reduction of batch sizes. The customer orders more frequently and in smaller quantities. The delivered quantity per order is reduced to a minimum of one position and one piece, which results in a disproportionately higher amount of work for the supplier’s order-picking systems. Therefore, the supply inventory is shifted from the buyer to the seller.
The following changes caused by e-commerce are to be considered (Seemüller 2006):
- Increasing rate of change in the variety of articles
- Volatile order batch size and rapidly fluctuating throughput demands
- Increasing volume due to more and more people ordering over the in-ternet
- Fewer positions per order. This is a phenomenon which is caused especially by e-commerce
- Strongly fluctuating workload of the systems within a single day, over a period of several working days, as well as during the course of a year (e.g. Christmas business)
Because of these factors internal material flow systems have to become increasingly flexible. Inflexible systems like conventional conveyor belts are being replaced with more flexible systems, and instead of large scale systems, modular, scalable solutions are increasingly in demand. Existing flexible systems (forklifts, hand forklifts, picking carts, or hanging cranes) are uneconomical due to high labor costs and low throughput rates, and are only partially suitable for the transport of smaller conveyor units. To avoid the current compromise between flexibility and automation, plant manufacturers are increasingly trying to build modular plants that can be set up in the shortest amount of time in accordance with the demands of the customer. These modular components must still be installed and programmed individually, since no two conveyor systems are the same and the conveyor components cannot electronically integrate themselves automatically into the overall information system. A future change or expansion therefore requires a large investment, whereby a reutilization for other products or processes is not economically feasible.
Even if newer components already possess their own motor and sensor controls, the overall organization of the material flow elements is done centrally. The necessary programming effort negates the big advantage of modular systems, namely the simple, mechanical construction of the complete system by linking together a small range of standard components. Up to now this was necessary, because a centralized system architecture that controlled the system as a whole was used, which needed to be adapted to each individual layout at the time of installation or when changes were made. In spite of modular systems, long project development times, high costs for experts for the installation and modification, as well as long downtimes for repairs remain.
The continuous development of basic technologies has contributed to the fact that the prices of mass-produced items like CPU and memory chips, sensors, and identification systems are continually decreasing. This development allows thoughts about completely decentralized material flow systems where the individual components are equipped with all necessary electronics, thus allowing the organization of the material flow to be distributed among the components. Therefore, not only modular material flow systems can be developed, but also more autonomous and intelligent conveyor systems where each module takes over a part of the transportation task and exchanges information about the current system status with other modules. The immense increase in flexibility and the reduction of investment costs by avoiding the current programming and planning costs would especially enable small and medium-sized companies the increased use of automated systems. A change of topology during operations would be just as easy as the replacement of single components in case of a malfunction.
Nevertheless, the biggest challenge of decentralized systems is the lack of an overview and with it the complex coordination of the conveyor units, so that no collisions or deadlocks can occur, especially during periods of heavy utilization.
1.2 Aim of the dissertation
The aim of this dissertation is to establish the foundation for the future application of completely decentralized material flow systems. At first, existing decentralized systems such as IT networks are analyzed and the structural differences in comparison to the transportation of physical conveyor units are established. An overview of existing approaches to the decentralized control of material flow additionally illuminates the latest knowledge and previously implemented concepts.
After defining the basic requirements and a clear range of functions, a new, completely decentralized control system for a continuous conveyor system is developed that distinguishes itself from all previous systems through a higher degree of autonomy and decentralization, as well as being able to operate completely independently from the infrastructure. Furthermore, all conveyor sections should be able to transport conveyor units in both directions.
Through the implementation of the control algorithms in a simulated environment, their performance is tested to determine their characteristics in relation to throughput and deadlock handling. The focus thereby is placed first and foremost on the stability of the system, so that in the worst case the flow rate will be reduced, but the system will remain functional at all times and a deadlock can never occur. Also, in the case of a technical failure of a module, the system should react appropriately. The knowledge gained should serve to prove the operational reliability.
The technical validation of the system occurs through the design and construction of several independent modules that are able to transport conveyor units, e.g., small load bearers (SLBs) or pallets, in accordance with the established basic requirements. At the same time, the control principle developed here is implemented in the modules and all functions are tested under laboratory conditions.
1.3 Structure of the dissertation
Chapter 2 ”Centralized material flow control” gives an overview of conventional material flow controls. Following the classification of different types of controls, the most common hardware control - the programmable logic controller (PLC) - is discussed in detail and its mode of operation is examined. In conclusion, the limits of centralized material flow controls are identified to make clear the necessity to develop decentralized controls.
Chapter 3 ”Decentralized control systems” discusses currently existing, decentralized control systems. To begin, a possible definition of the term decentralization is undertaken to be able to classify the data management as well as the data processing in material flow systems in relation to their degree of decentralization. Next, the operation of IT networks is illuminated, since there are several similarities between the transport of data packages and conveyor units that serve as the basis for the new control concept. The physical topology and the transmission of data packets are discussed in detail. Furthermore, the prevention of collisions and the creation of routing information using the two fundamentally different procedures of Distance Vector Routing and Link State Routing play a role. Finally, the physical differences between the transportation of data packets and conveyor units are discussed.
Chapter 3 ends with an overview of ongoing initiatives for the development of decentralized material flow controls by focusing on the ”Internet of Things” and demonstrates the differences to this dissertation.
Chapter 4 ”Completely decentralized, autonomous continuous conveyor system” describes my own approach to the development of a decentralized control system for transporting conveyor units without any centralized infrastructure. After delineating the problem and defining the basic requirements, the decentralized approach to a control system is described in detail. Four individual steps of the controlling process that fulfill the material handling task of a single module are introduced. After the manual connection to the conveyor system, the modules must independently generate the necessary information about the topology of the system. Furthermore, they must recognize an arriving conveyor unit and be able to identify its desired destination (sink). After routing and reserving the transport path, the conveyor unit must be sent on its way.
The focus thereby is on the avoidance of deadlocks, which can occur above all in complex layouts. After explaining the control algorithm, the origination of a deadlock is addressed. It is shown how these deadlocks can be prevented, and also under which conditions a deadlock could still occur.
The last part of the chapter covers an analysis of the efficiency and capacity of the control system by investigating the throughput of various conveyor systems. The knowledge gained is ultimately used to develop a self-regulating mechanism that ensures that, even with a high injection rate of conveyor units, the system will consistently work at the highest possible filling rate.
Chapter 5 ”Technical implementation” introduces the first prototype of a module that is suitable for use in industry, and which, with the help of the control algorithm and upon being connected together with other modules, is capable of creating a material flow system with completely decentralized control. This chapter includes a discussion of the mechanical construction as well as the control components.
At the end of the dissertation, a summary of the gained knowledge and a short overview of further steps to the successful industrial application of a modular, decentralized control of material flow systems are given.
2 Centralized material flow controls
Until now, centralized control concepts have been used in material flow automation almost exclusively. Although there have been attempts to move more functionality closer to the actuators (e.g. the motor control mounted directly on the motor, which has an electrical power connection and a bus interface), the actual workflow logic still is in a central processing unit, which accesses the sensors and actuators via the bus system. However, it becomes clear that for the development of a decentralized material flow system, new demands on the peripheral hardware arise, which must be expanded to include the module for the workflow logic of the material flow.
In the following section, an overview of existing control methods is given. Afterward, the most commonly used control, the programmable logic controller (PLC), will be discussed in detail, in order to determine if existing hardware can meet the requirements of a potential decentralized application.
2.1 Classification of material flow controls
”Until today, control systems are used for the operative control of workflows. In the process, the relevant conditions are registered by sensors and transferred as input variables to the control system. The system interprets the input variables and forms output variables according to the given rules, which influence the real process with the assistance of actuators” (Jünemann and Beyer 1998).
2.1.1 Control concepts
Centralized controls
If the complete processing of the input variables of a system is done by an independent control unit that contains the complete logic, it is referred to as centralized control. Such a control unit could be a programmable logic controller (PLC) or an industrial PC. The capacity of the central unit is crucial for the entire system. Two identical units are often connected in parallel to increase availability. Such an arrangement can be used for applications that don’t place very high demands on the main computer and where a malfunction won’t cause excessive damage (see Figure 2.1a).
illustration not visible in this excerpt
Figure 2.1: a) Central, b) hierarchical, and c) decentralized control concept
Hierarchical controls
Hierarchic controls have a clearly defined, firm concept of the division of labor. A central control unit distributes tasks to group control units, which in turn manage individual controls. This concept allows more demanding tasks than centralized control, because some of the calculations and decisions are processed on a lower level, thus relieving some of the load on the central computer. Consequently, a breakdown of the central unit does not always lead to an immediate breakdown of the entire system because the subordinate systems can fulfill their tasks on their own for a certain time.
Decentralized controls
Each control unit communicates on an equal basis with all others. At the same time, controls can be functionally subdivided according to specific abilities. Here, the capability of the individual components is not decisive, but rather the capability of the communication system. In the case of pronounced modularity, maintenance is simpler and the system can continue to be partially operative while being serviced (Haaß 1997), (Langmann 2003).
2.1.2 Structure of the control systems
Hardwired controls
The program flow of the hardwired programmable logic controller is defined by the hardware and is only changeable within a very narrow range without changing components. It is mostly realized electronically but it can also be done hydraulically or pneumatically. This was the only possible control before the invention and distribution of programmable logic controllers (PLC). Due to the enormous amount of time and effort required for the wiring and the very low flexibility, these controllers are only relevant today in special applications, for example, when maximum security (e.g. emergency-off circuits, security fence monitoring) or speed is required. But even here they are increasingly being replaced with PLCs (Jünemann and Beyer 1998).
Programmable logic controllers (PLC)
PLCs allow a fast and flexible change of the process cycle because the corresponding control logic exists as a software program that can be adapted quickly. During the physical reconstruction of the plant, further expenses arise, in addition to the programming, for the wiring of the input and output signals to the controller. Using bus systems can reduce this expense.
Industrial PC (IPC)
Due to the massive drop in price and the rising dependability of personal computers, they are increasingly being used as an alternative to PLCs for machine control. Therefore, the sensors and actuators have bus interfaces and are connected directly to the PC. The PC contains the control logic and communicates with the higher level systems. Because to date most sensors and actuators do not have an integrated bus interface, several sensors are connected together electrically in switch boxes to form a bus client.
2.1.3 Signal processing
Synchronous control
Signal processing is synchronized with the clock signal. The ports are read, processed and written to cyclically. Most PLCs are using this form of processing because, due to the structure of the computer, it is easy to implement. However, attention must be paid that the cycle time is short enough for the application.
Asynchronous control
There is no clock signal. Each change in an input signal triggers a program sequence. This is realized, for example, with a relay control, which is not really very flexible because, being mechanical components, they can only be altered with a large amount of effort.
Logic control
With this method, the possible input signals are distinctly assigned to output signals using Boolean gates. The lack of storage elements is problematic, which is why a pure logic control is very rare.
Sequence control
The sequence control is a step-by-step procedure. The execution of the next step as required by the program depends on the stepping conditions. These can be conditional upon either time or a process. Time conditions, for example, allow a rest period for cranes to wait for the load to stop swinging. Process conditions use sensors to ensure that a load was correctly attached before it is set into motion.
These distinctive features cannot be assigned clearly to a certain control. Real controls are in fact a combination of the above-mentioned types as, for example, a synchronous logic control (Jünemann and Beyer 1998).
2.2 Tasks of the control levels in the material flow automation
In most cases, automated material flow systems are controlled at machine level with the support of a PLC. The problem here is logically connecting the data from the ERP (Enterprise Resource Planning) system, which controls the higher-level processes, to the PLC on the execution level. Current solutions are based on manufacturing execution systems (MES), which are located between the planning level and the execution level (see Figure 2.2). The EPR system plans the procedures and the MES carries them out. The MES processes the information from the ERP and generates the control signals. The communication occurs in detail as follows:
The MES receives transport orders from the ERP system and transforms them into orders with system coordinates for the controls level, because only these contain a detailed model of the layout. The system control, which can also be directly integrated into the MES converts the system coordinates into precise instructions for the PLC. Depending on the state of the conveyor system, the route of the conveyor units is optimized by the system control. Ultimately, confirmations of the successful completion of the operations or malfunctions are reported to the ERP system. New concepts allow the direct connection of the ERP system with the PLC through an extension of the PLC, whereby costs of central components are not incurred and decision authority is moved to a lower level (Arnold 2006), (Jünemann and Beyer 1998).
illustration not visible in this excerpt
Figure 2.2: Logistic core processes and IT levels, Source: Arnold 2006
2.3 Operating method of conventional PLC controls
The first programmable logic controllers (PLC) were invented in 1969. Since the 1980’s they have been very widely used in the industry as controllers of machines and mechanical systems, and have replaced the hard-wired arrangement of the relay technology. Presently there are approximately 300 companies in europe that offer PLCs for various applications. The largest worldwide suppliers are Siemens, Mitsubishi, Bosch-Rexrodt, FANUC, and Rockwell Automation (Allen-Bradley).
2.3.1 Field of application of a PLC
The fields of application of PLCs are extremely diverse. From the control of roller shutters in house technology to the linking of machine tools to controlling and monitoring of large-scale chemical plants, appropriate PLC designs are offered. Figure 2.3 shows the typical construction of a Siemens PLC, designed to be installed in an in-plant electrical control cabinet. Distinguishing features are in particular the number of input and output channels, the cycle time, the performance of the CPU, and the expandability. A further selection criterion is the safety requirements placed on the controller (Wellenreuther and Zastrow 1998).
2.3.2 Structure of a PLC
A PLC fundamentally consists of a power supply unit, a central hardware unit that contains the RAM (memory), EPROM (erasable, programmable, read-only memory), ROM (operating system), and CPU (central processing unit), and the input/output modules (see Figure 2.4). The latter contain the connections for the wiring and the actuators, like motors or valves. A PLC can be expanded easily by adding additional input/output modules. Many individually-wired connections can be replaced by bus systems nowadays. With the help of analog-to-digital-converter (ADC) not only digital but also analog signals can be processed whereby the control of mechanical drives is made possible. To increase reliability, the central components of a PLC can be integrated in parallel (Wellenreuther and Zastrow 2008),(Grötsch 2004)
illustration not visible in this excerpt
Figure 2.3: PLC (Simatic S7-400, Source: Siemens) .
illustration not visible in this excerpt
Table 2.1: Performance data of current PLC central units
2.3.3 Programming languages
Several programming languages are generally available. The programmer can also switch back and forth between languages during programming. In general, there are no advantages or disadvantages. The choice of language often depends on the personal preference of the programmer. The ladder logic or ladder diagram (LD) language, which is in practice quite popular because of its graphic troubleshooting support, is symbolically and logically closely related to an electrical wiring diagram. Instruction list (IL) is very similar to Assembler and is considered the most powerful language because, as opposed to LD, it recognizes more commands and allows jumps. A compromise between IL and LD is the function block diagram (FBD), which understands more commands than LD, but the diagrams are more clearly arranged than in IL. Structured text leans more toward higher level languages like Pascal or C. The programs can be written, adapted and tested on PCs in the office and transferred to the controllers (Pickhardt 2000), (Rolle 1998).
illustration not visible in this excerpt
Figure 2.4: Internal structrue of a PLC, Source: Fern-Uni Hagen
2.3.4 Operating concepts
The operating concept of the PLC is of central importance. Most PLC units work cyclically, whereby the process image of all inputs is captured in a type of infinite loop. These input variables are processed according to the stored program into output variables, which are then set as the process image of all outputs. Typical cycle times lie between 1 and 50 ms.
In connection with cycle times, the subject of realtime data processing should be mentioned. The term realtime taken by itself reveals nothing about the system capability. Realtime guarantees only a deterministic time response in reaction to an input signal, meaning the adherence to particular time limits under all circumstances. These limits have to be established on the basis of the actual process. In the automation and material flow area they usually lie in the millisecond range. A further demand on realtime systems is the concurrent processing of multiple tasks (computing processes). For this parallel computers or scheduling procedures (usually priority controlled) are used (Jünemann and Beyer 1998).
Event-controlled concepts, where all input signals are processed in sequence, are becoming increasingly popular. The advantage here is the guaranteed processing of each signal. The length of time until processing is however not exactly predictable. At best, when no events are waiting, then the reaction to a signal can be immediate and the controller is faster than a cycle-oriented controller. If the scaling of central systems is too large, then even these controllers reach their limits because too many input signals must be processed in a short time (Grötsch 2004).
2.3.5 PLC networks
Networking of multiple controls is common these days. This makes it possible to transfer the system states and now allows the automated linking of machines with robots, loaders and material flow systems. A frequently used interface is the so-called Profibus. It is a specific type of bus system with defined standards so that components from various producers can be linked together.
Communication networks
To allow communication between two controls, both sides have a modulator and a demodulator. The modulator converts the signal to be transmitted, which is in the form of a bit pattern, into a physical signal. This signal can be optical or electrical. The physical signal is decoded in the demodulator of the receiver. Details of the transmission parameters are given in the communication protocol, like, for example, voltage level and the length of time for the representation of a bit. Next to this relatively easy transmitting assignment there are many supporting services. These primarily avoid disruptions caused by other signals that are on the transmission path at the same time through access control to the channel (e.g. token ring). The function of a token ring is briefly described here because the Profibus, which is described below, uses this procedure to control access to its participants. The token is the permit to transmit. It is handed-off after a pre-determined time interval among all active participants in accordance with a logical sequence. The bus member who currently owns the token is temporarily the master of the network and is able to send requests to the other participants (temporary slaves). They answer the master’s requests but are not allowed to send any further data themselves.
The Ethernet, however, does not know this type of access control. The transmitter listens to the network and if there is no traffic, he transmits his data packet repeatedly until it returns to him correctly as an echo (Jünemann and Beyer 1998).
The same functionality of packet shipping and collision avoidance is used as the foundation for the decentralized material flow control and is therefore described in more detail in chapter 3.3.2.
Profibus
The Profibus interface is standardized and allows communication between PLCs as well as between the PLC and actuators and sensors. Thanks to bus technology the necessary wiring is minimal. Profibus systems are primarily used in manufacturing cells and assembly facilities, where they take over the communication between machines from different manufacturers.
PLC processes are usually executed asynchronously. Thereby a high variability in the process and reaction time emerges. With the help of the equidistantly-working Profibus it is possible to read all inputs with chronological synchronism, to evaluate them and then to set the outputs with chronological synchronism. This guarantees consistency of the data and adherence to the logical order. The complete control process becomes deterministic. As a result a reduction of the control time is possible because the data from the inputs can be provided ”just-in-time”. This means the inputs are already being read ahead of a new cycle and are available to the program directly at the beginning of the cycle. The outputs are handled in the same manner, in that they are written to at the beginning of the next cycle. A time advantage at the beginning and end of each cycle is the result. However, the number of periphery inputs and outputs increases the cycle time. The longest time of the single components determines the total time of the system. The more components that are connected, the worse the time response becomes. Above all, the performance of the CPU is decisive (Wellenreuther and Zastrow 2008).
Industrial Ethernet
The Industrial Ethernet emerges through the integration of a networked control into a company-wide network, which, for example, can create the connection of a material flow management system with the underlying controllers. Also, the Industrial Ethernet can enable a direct access to shared machine parameters through the Internet. Figure 2.5 shows an example of a graphic user interface of such a remote diagnosis.
Further applications are remote diagnosis units that are integrated into modern systems and allow fast help with a malfunction of the controlled machines by the manufacturer (Metter 2007).
illustration not visible in this excerpt
Figure 2.5: System diagnosis via the Industrial Ethernet
The capability of PLC processors and the size of memory modules are steadily increasing so that today, besides their main function, PLCs can also assume monitoring functions. The controls record the operating data of the machines and evaluate them. In addition to the graphical display on the human-machine interface (HMI) the data can be posted on the network and with suitable software tools, which are offered by many control manufacturers, made available to production planning and maintenance in realtime.
One concrete example of the Industrial Ethernet is the Profinet, which uses TCP/IP technology. The field bus integration describes the embedding of Profibuses in the Profinet. New kinds of transfer protocols (e.g., the IRT-mode) achieve cycle times that are shorter than 1 ms and therefore suitable for realtime applications. Motion control, i.e., the activation of actuators (e.g., CNC axes of a machine tool), is also integrated into the network. The connection of decentralized field devices (e.g., starter motors, input and output components) is handled in the same manner as are components of distributed intelligence using Profinet. An extensive security concept is available with Profisafe, which minimizes the loss of transmission data through, among other techniques, consecutive numbering of data packets, monitoring of the exact transmission time, and identification verification between the transmitter and the receiver. Remote diagnosis and web integration complete the functional spectrum (Pigan 2008), (Messerschmidt and Lüder 2002).
2.3.6 Limits of and alternatives to centralized controls for material flow systems
The PLC owes its prevalence to its stability and simple programming. For a long time PLCs were not allowed to be used in safety-critical areas because it was feared that software or hardware errors (e.g., integrated circuit diffusion) could endanger peoples’ lives. Programs with built-in test functions, concepts like Profisafe and multi-channel systems alleviated the problem. Today, limits are primarily found in the areas of highly specific applications, for example, when extremely fast data processing and reactions to input signals are required. In these cases hard-wired systems are still used.
Because of the increasing flexibility requirements with regard to the adaptability of material flow systems, the programming effort involved with changes is moving more and more into the foreground, because a PLC always has to be readapted to the individual layout of the sensors and actuators. The problems thereby intensify when software updates are brought to the market in ever shorter time spans and it becomes more complicated to change the existing programming. Often it is hard to tell what impact a local change in hierarchical programming will have in other areas, because the complexity and the cross references become unmanageable in larger systems.
Another limit of conventional material flow systems lies in the restricted operational capability of existing systems. So far no conveyor system exists where all transport paths can be used in both transport directions. A change of transport direction is only possible with a large investment effort. A joint use of track sections in rail traffic, for example, are isolated, special-purpose solutions that until now have not been emulated in intralogistics.
Decentralization demands an increase in the functional range of individual components and the definition of new interfaces on a higher logical level. This simplifies the physical alteration of the transport system because necessary sensors and wiring remain consolidated with the material flow means. It is thereby necessary, however, that the control interfaces as well as the physical interfaces define a closed functional range and a ”plug-and-play” is created. This can be very well realized with existing PLC controls because processing power is sufficient for the additional logic for the coordination of the material flow. However, it could become necessary to install additional interfaces for this coordinating function. Alternatively, the PLC could also be replaced by small IPCs or simple computers, like, for example, microcontrollers. In this case, it must be ensured that the same programs are implemented on each individual, decentralized component and thereby reprogramming is avoided when the transport system is altered.
3 Decentralized control systems
3.1 Definition of the term ”decentralized” in material flow
In the course of the increased flexibility demands of the industry on material flow systems, the efforts to develop decentralized material flow controls have increased sharply. The goal is to bring more and more decision-making competence closer to the actuators and also to implement related functions physically into a compact unit containing mechanics and electronics. These units can then be connected together with minimal installation effort to produce the desired material flow system (Günthner and Wilke 2002).
Several products have already been developed further following these requirements. Sensors now not only convert physical measurements into a voltage level but some also have a bus connection that sends the interpreted value to the logic component (e.g., PLC or IPC). Motors are increasingly being equipped directly with their corresponding control circuits so that to control the motor, only information about speed, ramp or end position are necessary (Messerschmidt and Lorentz 2002).
Although the products have a decentralized character, they are still far away from being able to fulfill complex conveying duties independently. Nevertheless, these systems are frequently described as ”decentralized”. On the other hand, various research projects are already working on giving conveyor units the ability to independently find their route from source to sink and to be able to call available conveyor means like a ”taxi”. Such a system would certainly have a higher level of decentralization than just a bus-connected motor (Bullinger and ten Hompel 2007).
Therefore, a definition and a classification of decentralization is introduced below that should help to classify the development initiatives and system solutions better.
Assignment of tasks
The control of a material flow system consists of various components that fulfill different tasks. Figure 3.1 shows the assignment of tasks on multiple levels on which varying information and logical cycles must be processed. The degree of decentralization of such a system is determined by the proportion of data storage and data processing that is situated locally, that is, near the sensors and actuators.
illustration not visible in this excerpt
Figure 3.1: Distribution of control tasks in material flow systems on multiple levels, Source: Furmans and Arnold 2006
In modern intralogistic systems, instructions are generated by the central computer, which product (what) must have arrived at what point in time (when) at which destination (where). Push systems generate these instructions ”downstream” along the production process via the central computer, whereas pull systems transmit these instructions ”upstream” from each local position to the upstream station. Consequently, pull controls work with a decentralized provision of information but the processing is mostly done as before centrally via the central computer, and only the triggering of an order occurs locally.
Therefore it is suggested to measure the degree of decentralization by determining who owns and processes the three data of ”what”, ”when”, and ”where” within the material flow control.
Additionally, which entity has which part of the work flow logic for the completion of the transporting task must be analyzed. The work flow logic is mainly occupied with the question of ”how”. This question can be answered with the following sub-processes:
- generating of the topology information (path finding)
- routing (path choice)
- actuator control (execution of the transport)
- supervision (monitoring)
Figure 3.2 shows the structure of data storage and processing, which can be handled centrally by the material flow computer, or locally by the conveyor, as well as by the conveyor unit itself.
illustration not visible in this excerpt
Figure 3.2: Centralization vs. decentralization in MFC
The classification helps to distinguish whether the information about the identity of the conveyor unit, the destination address, and the desired time of arrival at the destination is stored at the conveyor unit itself or is centrally stored. A decentralized storage of this information by the conveyor system would make little sense because the conveyor system is only interacting temporarily with the individual conveyor units and therefore shouldn’t save any short-term, conveyor unit-specific data.
On the other hand, the processing of conveyor unit-specific data can very well take place at the conveyor system because a conveyor unit only has little information about the mechanical characteristics of the conveyor system and would therefore require additional effort be able to find the route through the conveyor system and operate the appropriate actuators by itself.
3.2 Research projects in decentralized material flow controls
Flexible material flow systems are currently the subject of many research projects. The ideas range from increasing the flexibility at the organizational level, to the further technical development of transport components, for example autonomous, driverless transport systems (DTS), all the way to the idea of the ”Internet of Things”, in which conveyor units autonomously find their way from production to customer - and back again to recycling. (ten Hompel and Nagel 2008).
In the area of the organizational level, the SFB 467 should be mentioned, which is concerned with ”versatile company structures for the multi-faceted serial production” (Westkämper, Wiendahl, and Balve 1998). The goal hereby is the development of models, methods and processes to increase the versatility of manufacturing companies. The focus here is on the versatility of the processes and less on the control and techniques of material flow systems.
A large volume of literature exists about segmentation (Wildemann 1988), holonic (van Brussel and Valckenaers 2000), or fractal factories (Warnecke 1995), (Wiendahl 2005).
During the planning and design of production facilities, the demand for an increase in flexibility is answered by the development of modular structures. (Schenk 2002) calls for this throughout the entire intralogistic chain, which ranges from conveyor technology, through storage techniques, all the way to sorting and order-picking systems.
According to (Jünemann and Schmidt 1999) conveying systems are evaluated according to various criteria as, for example, the layout flexibility and the throughput. Comparing these criteria shows a gap for highly flexible conveying means with high throughputs (see Figure 3.3). A part of this gap is to be closed by this dissertation.
illustration not visible in this excerpt
Figure 3.3: Classification of conveying systems for SLBs (< 50 kg)
To develop new concepts for flexible conveyor systems some research centers are engaged in the further development of conventional conveyor technology for a more flexible application. All initiatives are aimed towards a step-by-step decentralization of the control technology and the creation of flexible interfaces. The principle of the black box is used to modularize existing technology and to make it versatilely deployable, following the building block principle. Each technical conveyor component is thereby precisely described according to its input and output parameters, as, for example, electrical power supply, control parameters, payload, geometric dimensions and so on, without going into detail about what happens inside the components. This way, different components can be configured together into a complete conveyor system (Wilke 2006).
3.2.1 The Internet of Things
”The procedures for the control of (data) packets within the decentrally structured Internet are known and have proved to be efficient enough to manage the information and communication requirements of the earth’s population. This accomplishment qualifies the idea of the ”Internet of Things” to become the guide for the revolution of the material flow controls” (ten Hompel and Nagel 2008).
In the area of information processing, the ”Internet of Things” refers to mostly wireless, self-configuring networks between objects as, for example, home appliances. The first concepts were developed in 1999 at MIT in cooperation with the Auto-ID Center.
In the past few years the best-known initiative for the research of decentralized material flow systems has been advanced at the Fraunhofer Institute for Material Flow and Logistics (IML) in Dortmund under the direction of Prof. Dr. Michael ten Hompel. Under the same name ”Internet of Things”, concepts have been developed on how individual objects of a material flow system (for example conveyor units and conveying means), can achieve full autonomy in seven steps. The goal is to be able to transport goods completely autonomously, without any central control. The basis is provided by equipping all objects with RFID tags, which contain all the information necessary for the accomplishment of the task (Bullinger and ten Hompel 2007), (ten Hompel and Corban 2004a).
illustration not visible in this excerpt
Figure 3.4: Continuous conveyors with decentralized control at the IML for researching the ”Internet of Things”, source: ten Hompel, Libert, and Sondhof 2006
Decentralized control of a piece goods conveyor system
A continuous conveyor system serves as a demonstrator that consists of 36 conveying segments and is controlled with the help of 50 drives and 60 sensors. The system is equipped with 7 IPCs that can simultaneously run a separate control program for each of the 36 segments, each program customized for the respective segment. These were connected via the Ethernet (see Figure 3.5).
illustration not visible in this excerpt
Figure 3.5: Distribution of the continuous conveyor system in 36 segments, source: ten Hompel, Libert, and Sondhof 2006
Four RFID read/write devices with 868 MHz technology were installed and all transport containers were equipped with appropriate tags for identification.
The control program consists of two parts. The logical part was identical for all segments because it was supposed to be executed independently of the hardware. The second part was adapted to the hardware of each segment including the necessary I/O drivers. Because of this splitting, the generation of the individual segment controls was greatly simplified. To make a decentralized decision regarding the path of a container, the control programs have to know the topology of the system. This was manually produced and stored centrally over the network on a workstation PC as an XML file. This is read by the segment controls during start-up and a routing table is locally generated.
[...]
-
-
-
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
-
¡Carge sus propios textos! Gane dinero y un iPhone X.