Available online at www.sciencedirect.com ScienceDirect Procedia CIRP 57 (2016) 207 – 212 49th CIRP Conference on Manufacturing Systems (CIRP-CMS 2016) Industrie 4.0: Using cyber-physical systems for value-stream based production evaluation Erdal Tantik*, Reiner Anderl Department of Computer Integrated Design, TU Darmstadt, Otto-Berndt-Straße 2, 64287 Darmstadt, Germany * Corresponding author Tel.: +49 (0) 6151 16-21797; fax: +49 (0) 6151 16-21793 E-mail address: tantik@dik.tu-darmstadt.de Abstract The increasing application of cyber-physical systems (CPS) in manufacturing processes highly avails communication between production systems and products One cornerstone of these cyber-physical production systems (CPPS) is the digital representation (DR) of each product The digital representation contains significant product-specific information about all phases of the production process This information is mainly used to coordinate the production by altering and adapting the planning data This paper introduces a comprehensive method to examine and evaluate a production system using the information of the digital representations By including order-specific information, such as the product price and production costs, a value stream based evaluation of each production stage can be generated and visualized with very little effort The proposed modelling technique for machines, storages and transportation systems enables the application of this concept for highly automated production systems as well as shop fabrication A modified method of the comprehensive value stream evaluation is applied for a use case, showing the general procedure and advantages © 2016 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license © 2015 The Authors Published by Elsevier B.V (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of Scientific committee of the 49th CIRP Conference on Manufacturing Systems (CIRP-CMS 2016) Peer-review under responsibility of the scientific committee of the 49th CIRP Conference on Manufacturing Systems Keywords: Industrie 4.0; value stream; simulation; production system; evaluation Introduction To meet the increasing competition on global markets, advanced methods of the information technology and especially network technology are used in the production process [1] The impact of these technologies is imminent, thus it is seen as the Fourth Industrial Revolution [1,2] For this purpose the German federal government supports a high-tech strategy called Industrie 4.0 One facet of this technology is the interconnectivity of all components to monitor and control the production process The integrations of computation in physical processes are called cyber-physical systems [3] An increasing application of cyber physical systems (CPS) in the production systems leads to cyber physical production systems (CPPS) One key aspect of a CPPS is the amount of information gathered either continually or between important production steps, allowing a very precise comprehension of the whole production structure [4] This information is mainly used for the purpose of documentation and flow control of the production One aim is to realize a smart factory, meaning a factory which is aware of itself and the surrounding physical world, thus getting selforganized, flexible and allowing the production of customized products at low costs [5,6] But the gathered information upholds a huge potential to improve existing evaluation techniques While there have been few approaches to include additional product-specific data into the production planning [5], the general area is still unexplored This paper presents a concept to use the stored information for the evaluation of each production step individually and the production system as a whole Based on a comprehensive method for production system evaluation (CEM) [7], a modified method is introduced for the inclusion of additional data to improve existing indicators for production stations A simple generic production process is designed to verify the method 2212-8271 © 2016 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the scientific committee of the 49th CIRP Conference on Manufacturing Systems doi:10.1016/j.procir.2016.11.036 208 Erdal Tantik and Reiner Anderl / Procedia CIRP 57 (2016) 207 – 212 State of the art 2.1 Value Stream Mapping A very popular method for the optimization of production systems is the Value Stream Mapping (VSM) of the Toyota Production System, which is used to visualize the production process and minimize the “seven wastes” (inventory, transportation, motion, waiting, over-processing, overproduction and defects) [8] Therefore a predominant amount of research focuses on the optimization, standardization and modification of the VSM and other lean methods [7] Lately some approaches have been developed to apply value streams to the product development process However it has become obvious, that the techniques to enhance production and development processes have fundamental differences due to the interdependencies between both [9] But to create the VSM a huge amount of data about the real production process including scheduling, delays and information flows is needed, making the concept expensive in terms of time and effort Furthermore, the VSM is only suitable for one product or product family [5], thus not being able to apply to flexible CPPS with highly customized products To meet some of the problems, especially to include additional key performance indicators and enhance the applicability, a comprehensive evaluation method (CEM) was developed [7] The overall target of these methods is to demonstrate the impact of alteration by comparing two states of the production system Therefore the results merely indicate whether an alteration has improved the process, but not where the flaws of both system are located Consequently one approach for a new optimization method is to modify an approved method to meet the flaws Another requirement derives from the huge variety of evaluation and optimization aspects of the production system, new methods ideally have to cope with While the CEM is relatively new, one advantage of this method is the ability to include an arbitrary number of indicators with highly adaptable characteristics Those indicators can furthermore be prioritized, which increases the application scenarios, making it particularly suitable for the new method The DR contains information about real and planed phases of the entire product life cycle Therefore the bidirectional communication between the DRs, the physical components and production system can be used to control the production flow and highly adapt the production process to the current conditions, leading to an increased flexibility without lowering the productivity While the data is currently used for the main goal of a real-time production control, the gathered information furthermore can be used to examine and evaluate the production process The integration of DR-data into the evaluation method reduces expenses and by using real production data the evaluation results improve in terms of precision Components and structure To meet the huge variation of production systems a generative structure for the process entities is needed The main entities can be generalized e.g in buffers, workstation, stocks and transport activities which can be further subdivided, thus enabling to define the scale of each entity and a strong simplification of the production process For the purpose of the following use case a very generic structure was chosen, whereby the structure of every production system can be represented by the same entities, seen in Fig There are only two block types; workstations and stocks The workstation has to represent every activity including processes, transportations and inspections Every workstation itself may consist of substations, allowing a verification of different hierarchical levels of the production By only comparing entities of the same level, this structure has huge capabilities to combine and compare different production areas 2.2 ICDM and evaluation data One key aspect of CPS is the ability to use integrated component data models (ICDM) Instead of using one generic data model for every similar component, named integrated product data model, with ICDM an individual digital representation (DR) is made for every real component [10] Fig Generic representation of a production system with two different entities Each order is divided in batches Every time the batch changes the workstation or stock, the process data is stored By including order-specific data, such as product cost and profit, the exact value stream within the production process can be described Evaluation technique based on the value stream Fig Exemplary data content of an ICDM [10] The main part of the CEM is a five step procedure from identifying the key performance indicators (KPIs) to calculating the overall benefit of a scenario To meet the requirements the procedure was modified as depicted in Fig While the first three steps remain identical in terms of identifying and prioritizing key indicators, the context changes 209 Erdal Tantik and Reiner Anderl / Procedia CIRP 57 (2016) 207 – 212 from production system to single sections of the production process i.e production cells and activities Fig CEM procedure including the modifications indicator of its own, reducing the flexibility and complexity of the application The third way is to apply the evaluation method and consider the parameter subsequently This separation allows an evaluation approach of different parameters without their interactions Every key indicator should be defined to summarize one aspect of each entity, e.g the flow degree or capacity Furthermore by using the CPPS and DR-data, it is possible to locate the main value paths within a production system This is especially important for high product variation and customtailored products For the value flow, the indicator can be composed of the monetary value of the batch after the production process minus the initial value This information is commonly known before the production starts In many cases additional parameters, such as intermediate values of the batch can be approximated and therefore factored into the KPI For assembling processes this is more difficult A simplification is to assume that the value of the single component is proportional to the last known value before the assembly section The first step is the identification of the KPIs, which can differ significantly depending on the entity The next step is the mathematical definition of the impact of every KPI The definition normalizes a simple value function (SVF) of the key indicators resulting in a value of zero for the worst possible and one for the best possible outcome The shape of SVFs is not predefined (see Fig 4) Fig Exemplary section of an assembly process with known batch values A section of an assembly process can be seen in Fig The batches with value va and vc are assembled in station With the assumption that all values are increasing proportional to the last known value, the value stream from station to station can be approximated by: Fig Setting up a simplified value function (SVF) of the adapted key performance indicator with a customized shape Contrary to the original approach, the key indicators are not parameter of the overall process states but of the individual sections within one production step Thus the partial benefit un,z of the CEM becomes a simplified value for the current production section Depending on the entity structure, different KPIs and weightings can be used, thus enabling the consideration of all characteristics of each entity In the last step, all station of the same level are compared to find the highest optimization potentials 4.1 Key performance indicators for different blocks A free definition of the KPIs in CEM enables a high applicability of the method but entails the scope for interpretation Three possibilities exist to consider a parameter for the evaluation The first one is to include the parameter in the mathematical definition of several key indicators Because the structure of the mathematical definition of the KPIs is not predefined, this allows to highly control the impact for different parameters but increases the complexity, especially by leading to mutual independencies between the value functions A second way is to define the parameter as a separate key v12 va (vm (va vb )) (va vc ) (1) In the same way v53 can be calculated For the evaluation it can furthermore be important to describe the rate each entity is holding up the overall production flow Due to the continual monitoring of entities and batches in a CPPS, all necessary data is available The target entity of each batch is known, thus it can be measured, how long the next block cannot receive the stored batch From the obtained time scales an indicator can be defined to describe, whether the current stock is holding up the process In practice this can be done by integrating a request and response system between entities Before moving a part from one entity to the next, a request for space is send to the target entity The response can be positive for sufficient capacities or negative and thus indicating the blocking of the progress The amount of negative responds from subsequent blocks ݎ௦ and negative responds to previous blocks ݎ are the main parameters for the hold-up indicator To reveal the main cause of the hold-up and find the entities with the highest potential to increase the material flow, both parameters of each block have to be compared (Fig 6) The higher the difference ݎௗ is, the more it is evident that the 210 Erdal Tantik and Reiner Anderl / Procedia CIRP 57 (2016) 207 – 212 current entity is blocking the process flow By including the number and value of the batches, the hold-up impact ݕ can be further specified Fig (a) request and respond count; (b) exemplary process holdup due to process Because stocks are mostly used as buffers without process time, they not send a request in every time step but only while not having sufficient capacities to receive batches from other entities Stock To rate stocks, the indicators for costs, fullness, capital lockup and hold-up should always be considered along with other user-specific indicators The first indicator includes monetary factors such as the current costs of the stock They can often be improved by decreasing the stock size Furthermore the fullness of the stock should be monitored to avoid unused capacities Additionally the products stored within a stock increase capital lockup These three indicators favour small stocks and highly flexible systems But very small stocks can block the process flow, thus the hold-up indicator ݕ should have a high weighting factor Vsk f ( yc , y fu , ylu , yh [, ]) (2) Workstation The workstation block itself is the actual value-producing entity and in many cases has the highest costs, thus offers huge optimization potential The operating cost of the workstation can hugely depend on the process flow This costs can be static or dynamic The dynamic costs consider the reduction during downtime For complex stations the monetary factors might furthermore be batch-dependent Other KPIs include the operating time, the capital lockup and the hold-up impact of the process and further user-defined indicators Vws f ( yc , y pt , ylu , yh [, ]) (3) Application and use case For an exemplary application of the developed method, a highly simplified production process was simulated The aim was to visualize the most critical production flow and furthermore to decide which products should be deleted to improve the process flow with little effect on the overall profit Therefore a basic discrete-event simulation software was developed To minimize complexity and retain the validity of the method, a fully automated production system is illustrated without any assembly processes The time is measured in native time steps (TS) Furthermore the monetary value of each batch is known The production systems consists of nine process stations, each having one stock as a pre-buffer and one as a post-buffer Each buffer has a capacity of three batches For the simulation 39 batches are processed The batches are categorized in three identical waves The first wave starts with the simulation, the second one starts after TS and the last wave starts after 10 TS All parameters of the batches are shown in Table The table lists the order time of the batch (not including the start time of the corresponding wave), the monetary value, the ranged ID of the workstations it needs to proceed and the number of TS it needs in the corresponding workstation With this information the simulation can be carried out Table Main parameters of batches in the use case batch ID order time value station ID station time 1,[14,27] 1, 2, 3, 4, 1, 1, 5, 10, 2,[15,28] 1, 2, 3, 5, 4, 1, 1, 15, 10, 20, 20 3,[16,29] 1, 2, 3, 5, 4, 7, 1, 5, 10, 10, 5, 5, 4,[17,30] 18 1, 2, 3, 5, 8, 1, 1, 5, 5, 20, 5,[18,31] 1, 2, 5, 4, 1, 5, 10, 20, 10 6,[19,32] 1, 2, 5, 4, 7, 1, 1, 5, 20, 15, 15 7,[20,33] 11 1, 2, 5, 8, 1, 1, 10, 20, 10 8,[21,34] 10 1, 2, 5, 7, 1, 1, 5, 5, 10 9,[22,35] 2 1, 2, 6, 5, 4, 7, 1, 5, 5, 10, 20, 20, 15 10,[23,36] 1, 2, 6, 5, 7, 1, 1, 15, 10, 15, 11,[24,37] 12 1, 2, 6, 5, 8, 1, 1, 10, 5, 5, 12,[25,38] 19 1, 2, 6, 8, 1, 10, 10, 20, 15 13,[26,39] 1, 2, 6, 5, 4, 1, 5, 20, 5, 20, 10 The task is to visualize the value stream and the flow degree of the production system For the calculation of the flow degree two different approaches are used The first method is to rate each block by parameters which can be easily gathered without the use of CPPS For buffers the main parameter is the average fullness and for the process sequence the normalized overall production time is used The second approach uses additional information gathered by the CPPS The evaluation is progressed according to the modified CEM The first step in the modified CEM is to identify the necessary parameters and their influence on the process flow To rate the workstation and buffers, additionally the capital lockup and the time of a workstation being blocked by the target entity is considered 5.1 Evaluation of the production system For every indicator a linear SVF is used Each considered indicator ݔhas a high value if there are few capacity, thus the highest value of the SVF is set at the point ݔҧ equals zero The blocked time of a workstation is the summarized time in which it contains a batch that cannot be passed on because of a full post-buffer The third step is the prioritizing of each indicator by a scoring system (Table 2) In this case it is not only important where a process is blocked but mostly how much value is blocked, thereby the flow for most profitable batches can be increased 211 Erdal Tantik and Reiner Anderl / Procedia CIRP 57 (2016) 207 – 212 Table Exemplary prioritization of key indicators in terms of the flow degree ylu ybt yf y pt sum rank g ylu - 2 42% ybt - 33% yf - 17% y pt 0 - 8% For the evaluation of the workstations the results of the SVF are multiplied with the weighting factor and summarized To visualize the overall performance, a modified Sankey diagram is used (Fig 7) Therein every workstation is depicted by one block containing the buffers The color represents the rating for the particular entity, thus showing where the most optimization potential is located In this case darker shades are used for an overall critical performance The width of the blocks and connectors between each block resembles the value stream each batch which shall be high The second parameter is the overall lockup of the batch, signifying how much capital lockup the batch produced in the production process This value should be very low The third parameter indicates how critical the batch is for the complete production flow and consists of three parameters: x the hold-up factor of the block: Batches proceeding a high number of blocks, which are causing the blockades have a low rating x blocked time: Batches passing entities, which are not being blocked have a high rating, because their effect on the overall flow is small x process time: Batches passing stations, which have high process time thus very low capacities have a low rating Using linear SVF and specifying the weighting factors, the equation for the evaluation value can be formed: Vb Fig Visualization of the production sequence (a) combined process time and fullness; (b) overall flow indicator For production lines, modified Sankey diagrams are suitable but for shop fabrication with very individualized products, generating arbitrary streams between the entities, the chord diagram is the better choice It can be seen in Fig 7a that S4 and S9 have very little capacities, thus blocking previous blocks Instead of focusing on the capacities, in Fig 7b the impact is considered by emphasizing capital lockup and blocked time S3 and S8 are most significantly affected by the poor production flow and thereby show high capital lockup due to their batch values It can be seen that in S8 the process is moderately fast but the post-buffer is unable to transmit the batches to S9 Because S9 has very few capacity, one way to reduce capital lockup would be to change S9 in such a way that it favors batches from S8 This could be done within S9 or by altering the start time of the batches The improvement of S3 is more complex because the process is too slow by its own and it is blocked by both, S4 and S5 5.2 Evaluation of the products The same method can be used to evaluate the products i.e batches In this case the aim is to calculate which products can be deleted to increase the production flow without strong effects on the overall profit For this purpose different indicators have to be defined The first indicator is the value of 0.5 y pf 0.25 ylu 0.25 ycr (4) By comparing the evaluation values, the worst performing batches can be detected In this use case the five worst performing batches have the IDs 28, 16, 15, 29 and 35 in that order In Table four relevant parameters for the production flow are shown for the initial state, after batch 28 was removed and after all five worst performing batches were removed The results show that the capital lockup and the overall production time of the whole system can be reduced significantly, improving efficiency and capacity of the overall production system without reducing the summarized value perceivably Table Production-flow relevant parameters before and after improvements process time lockup holdup initial 435 71426 316 without 28 change 415 -4.6% 70702 -1.01% 247 -21.8% without worst five batches change 370 63183 203 -14.9% -11.5% -35.8% value 324 322 -0.62% 308 -4.9% Summary and outlook The comprehensive value stream evaluation is extended to additionally evaluate different entities of a production system The results allow a comparison of different production conditions but furthermore indicate the location of the main flaws and highest optimization potentials By using a generative structure of stocks and workstations the method can be applied for different levels of the production process thus being suitable for complex production systems or manual work station alike The main advantages of the modified method is the ability to include supplementary data, gathered by cyber-physical 212 Erdal Tantik and Reiner Anderl / Procedia CIRP 57 (2016) 207 – 212 systems during the real production process, thus minimizing the expenses for the evaluation and analyses In further research projects this method can be advanced to include further KPIs commonly gathered in real production systems Another aspect is to refine the value stream by not including the profit or value of a batch, but the real value added by each production phase This would make it possible to not only visualize the main value streams in the production system but indicators containing the value addition relative to the costs Thereby the method could for instance help to decide, which processes should be done within the productions system and which ones should be outsourced, thus improving the optimization process Appendix Nomenclature g r v V x x y weighting factor rejected request batch value overall evaluation value dependent variable normalized dependent variable simplified value function Indices b c d cr fu h p pt lu ws s sk batch cost difference critical parameter average fullness hold-up previous process time capital lockup workstation subsequent stock Abbreviations CEM CPS CPPS KPI ICDM SVF DR VSM comprehensive evaluation method cyber-physical system cyber-physical production system key performance indicator integrated component data model simplified value function digital representation value stream mapping References [1] Dosch S, Spohrer A, Fleischer J Reduced commissioning time of components in machine tools through electronic data transmission Elsevier; Procedia CIRP 29 (2015) 311 – 316 [2] Schuh G, Reuter C, Hauptvogel A Increasing collaboration productivy for sustainable production systems Elsevier; Procedia CIRP 29 (2015) 191– 196 [3] Wang L, Törngren M, Onori M Current status and advancement of cyberphysical systems in manufacturing Elsevier; J Manuf Syst (2015) [4] Barthelemey A, Störkle D, Kuhlenkötter B, Deuse J Cyber Physical Systems for Life Cycle Continuous Technical Documentation of Manufacturing Facilities Elsevier; Procedia CIRP 17 (2014) 207– 211 [5] Schönemann M, Thiede S, Herrmann C Integrating product characteristics into extended value stream modeling Elsevier; Procedia CIRP 17 ( 2014 ) 368 – 373 [6] Picard A, Anderl R, Schützer K Controlling Smart Production Processes Using RESTful Web Services and Federative Factory Data Management 14th Asia Pacific Industrial Engineering and Management System, Dezember 3-6, 2013 [7] Sihn W, Pfeffer M A method for a comprehensive value stream evaluation Elsevier; CIRP Annals - Manufacturing Technology 62 (2013) 427–430 [8] Teichgräber UK, Bucourt M Applying value stream mapping techniques to eliminate non-value-added waste for the procurement of endovascular stents Elsevier; European Journal of Radiology 81 (2012) e47–e52 [9] Tyagi S, Choudhary A, Cai X, Yang K Value stream mapping to reduce the lead-time of a product development process Elsevier; Int J Production Economics 160 (2015) 202–212 [10] Picard A, Anderl R Integrated component data model for smart production planning DiK 2014; 9° Seminário Internacional de Alta Tecnologia ... change 41 5 -4. 6% 707 02 -1 .01 % 247 -21.8% without worst five batches change 3 70 63183 203 - 14. 9% -11.5% -35.8% value 3 24 322 -0. 62% 308 -4. 9% Summary and outlook The comprehensive value stream evaluation. .. 4, 1, 1, 15, 10, 20, 20 3,[16,29] 1, 2, 3, 5, 4, 7, 1, 5, 10, 10, 5, 5, 4, [17, 30] 18 1, 2, 3, 5, 8, 1, 1, 5, 5, 20, 5,[18,31] 1, 2, 5, 4, 1, 5, 10, 20, 10 6,[19,32] 1, 2, 5, 4, 7, 1, 1, 5, 20, ... 15 7,[ 20, 33] 11 1, 2, 5, 8, 1, 1, 10, 20, 10 8,[21, 34] 10 1, 2, 5, 7, 1, 1, 5, 5, 10 9,[22,35] 2 1, 2, 6, 5, 4, 7, 1, 5, 5, 10, 20, 20, 15 10, [23,36] 1, 2, 6, 5, 7, 1, 1, 15, 10, 15, 11,[ 24, 37]