1 Smart and Lean: A Case Study on Smart industry and the Process Flow
by
Thomas R. Pier
S2192454 t.r.pier@student.rug.nl
+31625347899
August 21, 2017
Master Thesis, Msc Supply Chain Management Faculty of Economics and Business,
University of Groningen, P.O. Box 800, 9700 AV Groningen,
The Netherlands
Supervisor: dr. Martin Land
Co-assessor: prof. Jan de Vries
2 PREFACE
I would like to use this preface to thank the people who contributed to the completion of this master Thesis. First of all, I would like to thank dr. Martin Land for all of his guidance, feedback and involvement. In addition, I would like to thank him for arranging my visit in China and for personally visiting me. Secondly, I would like to thank prof. Matthias Thürer for his inspiration and involvement. Furthermore, I would like to thank him for taking great personal care of me during my visit to China. I would like to thank everyone at the case company for all their time and efforts.
Additionally, I would like to thank the students of Jinan University for their help and translation efforts. Finally, I would like to thank prof. dr. Jan de Vries for his feedback during the research proposal phase and for his role as supervisor.
Groningen, August 2017 Thomas Pier
3 ABSTRACT
The concept of smart Industry covers interesting new technologies and concepts. However, little is known about the combination of smart industry, Lean and process flow. Whereas smart industry increases complexity through its data availability and new technologies, Lean desires to reduce it through variability reduction and simplification. This paradox is the subject of this research. We will research whether a combination of Lean and smart industry can be made to improve the process flow. We have performed a single-case study at a company in the manufacturing industry. We found an overall positive influence of smart on the process flow. This effect is both direct and through influencing various types of waste. We finalize our research by translating our findings into a design for a new smart improvement project.
Keywords: Smart industry, Lean, Industry 4.0, Internet of Things, Process Flow, Smart Components.
Word Count: 10.584 (excluding preface, references and appendices).
4 Table of Contents
PREFACE 2
1. INTRODUCTION 6
2. LITERATURE REVIEW 8
2.1. Lean and Automation 8
2.2. The Main Components for the Smart Factory 9
2.2.1. Smart Operator 10
2.2.2. Smart Products 10
2.2.3. Smart Machine 10
2.2.4. Smart Planning and Control 11
2.3. Process Flow 12
2.3.1. Planning and Control 12
2.3.2. Capacity Planning and Control 12
2.3.3. Inventory Control 13
2.4. Relation 13
3. RESEARCH METHOD 14
3.1. Research Design 14
3.2. Case Introduction 16
3.3. Data Collection 17
3.4. Design 19
4. RESULTS 19
4.1. Internal Warehousing and Logistics 19
4.2. Analysis of the current IoT system 21
4.3. Discussion 23
4.3.1. Smart and Process Flow 24
4.3.2. Lean versus smart 24
5. Design 26
5.1. Process Flow and Problems relating to Smart 27
5.1.1. Processes 27
5.1.2. Capacity 27
5.1.3. Quality control 28
5.1.4. Raw Material Management and Starvation 28
5.1.5. Synchronization 29
5
5.2. Smart solution 30
5.2.1. Quality Control 30
5.2.2. Raw Material Management 30
5.2.3. Synchronization 31
6. CONCLUSION AND FUTURE WORK 32
6.1. Managerial Implications 32
6.2. Future Research 32
6.3. Reflection 33
7. REFERENCES 34
Appendices 39
Appendix A. Smart industry integration model (Lee, 2015) 39 Appendix B. Articles involving Smart Planning or synonyms thereof. 39
Appendix C. Keywords used to find literature 40
Appendix D. Map of Production Plant A of Company X. 40
Appendix E. Interview and Observation Protocol 41
Interview Protocol 41
Observation Protocol 42
6 1. INTRODUCTION
This research focuses on smart Industry and how this can be used to improve the process flow according to Lean principles. More and more factories around the world are becoming smart, creating what we call Industry 4.0 (Zühlke, 2010). This comes from the fact that we are on the brink of the fourth revolution, the impact of which is expected to fundamentally change the way we work, live and relate to each other (Schwab, 2016). As industry is encountering this revolution, it raises the question which processes should be made smart and automated. Recent research found that the perfect level of automation for a Lean production system is when the flexibility of the production system is still able to adjust to customer demand (Harris and Harris, 2008; Hedelind & Jackson, 2008; Säfsten, Winroth & Stahre, 2007). Why is this necessary? As Taiichi Ohno (developer of the Toyota Production System) once said: “Forecasts are almost never correct, and they always change”. With this quote in our minds, the key to ever changing customer demand in a global market is flexibility (Harris and Harris, 2008). Furthermore, plentiful research has found that for companies to remain competitive they must be aware of new production systems and technological trends (Chen, 2010; Harris and Harris, 2008; Hedelind & Jackson, 2008; Orr, 1997; Säfsten et al., 2007).
The concept of smart industry covers interesting new technologies. These technologies are interconnected and this connection enables more complex smart industry concepts and components. Wireless manufacturing, a popular subject in recent literature, is a good example of this. It is enabled by technologies as the Internet of Things (IoT) and components thereof like RIFD and QR (Huang, Wright & Newman, 2009; Huang, Zhang and Chen, 2008;
Qu, Yang, Huang, Zhang, Luo & Qin, 2012; Wang, Wan, Li & Zang, 2015; Zhong, Dai, Qu, Hu and Huang, 2013). Allthese innovations lead to a much greater availability of data and the ability to adjust capacity instantly and at any moment. This enables the use of smart planning; being able to make decisions and adjustments to real-time sensory data from the shop floor (Huang et al., 2008). Recent research has found that smart planning improves processes in terms of flexibility, speed and reliability and allows for optimal use of such innovations (Huang et al., 2008; Kohlberg & Zühlke, 2015;
Lappe, Veigt, Franke, Kolberg, Schlick, Stephan, Guth & Zimmerling 2014;Zhong et al., 2012). Another possible implementation of smart Industries comes in the combination of smart and Lean, named Lean Automation. The combination may allow for an increase in flexibility and reduction in inventory (Kohlberg & Zühlke, 2015). The concept of combining automation with Lean is not new. It was first coined back in 1988 by Ohno, who called it Autonomation. This focused on production to have the capacity to stop automatically in case of irregularities.
Smart industry developments have been suggested to facilitate complicated planning issues, such
as online optimization and the use of IoT (Kolberg & Zühlke, 2015; Qu, Lei, Wang, Nie & Chen,
7 2015; Wang et al., 2015; Xu, 2011). Contrarily, the principles of Lean have taken the viewpoint of improving flow by reducing variability and complexity (Kolberg, Knobloch & Zühlke, 2016), which enables the use of simplified planning and control approaches such as Kanban (Lage Junior &
Godinho Filho, 2010). It seems these two concepts are contradictory because Lean is trying to reduce variability and complexity and smart industry seems to increase complexity; for instance, through its much greater data availability and the need to handle this data or its new technologies and the implementation of these technologies. However, making the right trade-off between the two may be the answer; trying to reduce complexity in situations that allow for this reduction on the one hand and making highly complex processes that can’t be simplified smart on the other. Given the current paradox between Lean and smart industry, this research will investigate the relation between Lean, smart industry, and the process flow. Current literature has suggested the need for such research. Zuehlke (2010) suggested that new technologies may be able to help a company become Leaner. According to their research the new type of technologies can avoid unnecessary technology and information. They also found that improved communication and fault detection may reduce waste. Furthermore, Jung, Morris, Lyons, Leong and Cho (2015) suggest the need for research on how the current challenges that smart Industry components bring to the process flow can be mitigated. A challenge the authors foresee is a manufacturer’s ability to translate internal requirements such as stock to external parties such as suppliers, because the internal planning thereof is now automated. Zhang, Zhang, Wang, Sun, Si and Yang (2014) suggested future research on production decisions and the influence of RFID technology. Also, Zhong et al. (2013) stressed the need for research on the use of planning and scheduling based on wireless real-time data.
This research will investigate whether smart Industry components can instead be used to improve the process flow while taken Lean Principles into account. We will furthermore shine some light on the benefits that smart industry components may bring. As presented above, the gap in current literature and the paradox in Lean combined with smart Industry leads to the following research question.
“How can Smart Industry developments help to improve the process flow, following Lean principles?”
In order to perform this research, we first provide a literature review. Next, we will discuss our methodology with the selected research method. We will then present the results of our case study. The selected case company has already implemented a smart system in a part of the process flow and we will use it to research our propositions. We will then discuss our findings.
Additionally, we will use our findings to present a design for a new smart project. This is followed
by the conclusion, recommendations and limitations.
8 2. LITERATURE REVIEW
A literature review is performed to better understand Lean, smart industry and the relation of these two to the process flow. We start by discussing Lean and its relation to automation. Then the smart factory components and technologies along with potential improvements are covered to gain a better understanding of what a smart system entails and what benefits it may have. Additionally, we will discuss the process flow in order to gain a better understanding of the concepts and processes relevant in our research. We will finalize the literature review with the relation between these concepts, creating propositions following the literature.
2.1. Lean and Automation
The Lean manufacturing approach strives to be able to meet demand in quality, quantity, time and location at minimum cost (Slack et al., 2010: 470). We will operationalize Lean as the production of goods with minimal buffering costs (Hopp & Spearman, 2004: 144). This means improving flow by eliminating obvious wastes (e.g. excessive set-up times or the use of unreliable machines), reducing inventory buffers and reducing variability in order to minimize time, capacity and inventory buffers. Finally, this cycle can be repeated in order to achieve continual improvement (Hopp &
Spearman, 2004). Effective Lean production systems should be a combination of both manual and automated processes and therefore the task is to determine the appropriate degree of automation with human involvement (Groover, 2000; Harris and Harris, 2008; Hedelind and Jackson, 2008;
Jackson, Hedelind, Hellström, Granlund & Friedler, 2011; Winroth, Stahre & Säfsten, 2006;).
Processes such as exception handling and logistic tasks are processes that are not (yet) reasonable to automate (Bilberg and Hadar, 2012). However, automation can have significant benefits such as negotiating cycle times and finding an optimum between the highest possible capacity utilisation and a continuous flow of goods (Kolberg & Zühlke, 2015). By doing so such a system would be able to improve flow. For instance, by optimising based on factors in Little’s law;
the current Work in Process, the cycle time and the throughput rate (Little & Graves, 2008). This would allow for a Leaner production system. Such an automated system will rely on connectable sensors (like RFID and QR codes) and an overall increased computing power of the system through cloud computing (Luo, Wang, Kong, Lu & Qu, 2017; Schmitt, Permin, Kerkhoff, Plutz &
Böckmann, 2017). This increased connectivity and computing power supports fault-prone processes (Kohlberg & Zühlke, 2015). Furthermore Orr (1997) found that automation can lead to cost reductions and an increase in production throughput with the existing manufacturing resources.
The concept of combining smart industries and Lean has been briefly discussed by recent
literature (Kolberg et al., 2016; Lee, 2015; Yoon, Shin & Suh, 2012). Yoon et al. (2012) have
9 published a conceptual factory (based on empirical research by Lucke, Constantinescu &
Westkamper, 2008; Meyer, Främling & Holmström, 2009 and Zuehlke, 2009) on how smart Industry components such as RFID, wireless network infrastructure and connected resources can overcome regular Lean manufacturing problems. They found that problems such as inaccurate demand forecasts or trouble with individualized production control can be overcome through an intelligent, adaptive, wireless manufacturing system. This leads to increased flexibility and efficiency, prior risk detection, and cost reductions. However, these researchers suggest a whole new manufacturing paradigm should be developed (Yoon et al., 2012; Lucke et al., 2008; Meyer et al., 2009; Zuehlke, 2010; Zuehlke, 2009), whilst other researchers suggest improvements on existing systems (Kolberg et al., 2016; Kolberg & Zühlke, 2015). The smart Factory contains technologies that differ from traditional manufacturing systems. In the next section, we will discuss the smart Factory and its components more in depth.
2.2. The Main Components for the Smart Factory
A smart object consists of two levels; the passive physical object level (e.g. RFID Reader) and an
active physical object level (e.g. RFID Tag) (Qu, Lei, Wang, Nie, Chen, & Huang 2016). Following
Kolberg & Zühlke (2015) these smart factory’s components can be split in four main groups: smart
Products that contain their own production information and communicate with smart Machines,
Smart Planning systems that can adjust instantly and allow for optimization and finally smart
Operators, humans with modern technologies that allow them to supervise operations. The smart
Planning can be seen as the glue between the smart components and at the same time as the
controlling mechanism above the components. As such it is enabled by these smart objects. The
smart planning also exchanges information, such as product or customer information, with the ERP
system (Figure 1). The integration of these components in the organization can be measured by a
model created by Lee (2015) (Appendix A), this model defines the smart Factory in 5 different
levels, with each level progressively becoming smarter. The most important indicator of these
levels is how the organization uses the collected data. As such the model focuses on the attributes
and functions of the smart Industry components compared to a traditional factory; in other words,
how a company actually uses the smart factory components to their advantage. We will further
discuss potential advantages of the smart Industry components and applications found in current
literature.
10
Figure 1. Smart Factory Components
2.2.1. Smart Operator
New technologies allow employees to become smart themselves. Using innovations like smart watches employees are able to receive information instantly, for instance error messages and the location thereof. The smart Operator will be able to improve the time from the detection of failures to the notification thereof. This helps a company become Leaner as it allows for a reduction in waste like downtime, rework and scrap (Ohno, 1988: 58). Additionally, the use smart tablets, smart phones and machine human interfaces allow for operators to exchange complex information with each other and the system (Luo et al., 2017).
2.2.2. Smart Products
Smart products are able to collect and contain data themselves during and after production. This will allow for less labour-intensive production and more accurate data (Kohlberg & Zühlke, 2015).
The accurate data in turn allows for improvements. Furthermore, smart products can contain information (e.g. Kanban information) to control their own production processes through technology as IoT through RFID chips and QR codes (Huang et al., 2008). These can enable real-time traceability and visibility of materials (Zhong et al., 2013) and can substantially improve shop-floor management and specifically WIP-handling (Qu et al., 2012).
2.2.3. Smart Machine
Technical installations can help in preventing employees making mistakes (Ohno, 1988). Smart components can be integrated in processes that are vulnerable to failure or faults. Smart machines allow for more reliable processes and earlier failure detection (Lee, Bagheri, & Kao, 2015).
Moreover, the combination of smart machines with a smart planning enables predicting the remaining life of assets, which helps in maintaining a JIT maintenance strategy (Lee et al., 2015).
Furthermore, smart machines support a flexible modular production. The German Research Centre
for Artificial Intelligence (DFKI) demonstrated modular working stations that can be flexibly
11 reconfigured to new production lines. This allowed the setup time for a whole production line to be significantly reduced, following the Single-Minute-Exchange-of-Die principle (Zuehlke, 2010).
These results are in line with research by Bieringer, Buchholz & Kockmann (2013) who found that modularity may be the answer to current problems in the chemical industry, such as long setup times and inefficiencies resulting in an increased flexibility. Another possibility of smart machines is the integration of sensors. For instance, by integrating a camera in bins, it is possible to measure inventory and if necessary, order new supplies automatically (Würth Industrie Service GmbH & Co.
KG, 2013). The use of these sensors also prevents mismatches in the inventory of the system versus the real inventory. The introduction of these technologies with allow for a reduction in buffer stock, simplify the high manual effort of maintaining stocks, and increase overall response speed (Schuh, Stich, Reuter, Blum, Brambring, Hempel & Schiemann, 2017).
2.2.4. Smart Planning and Control
The term smart planning is ambiguous in the current literature about Smart industry concerning both the technologies involved and the term they use (see Appendix B for details). In this research, we will define smart planning as a planning system that uses wireless real-time data, that can in turn translate that data to meaningful information and successively a person or system is able to adjust operation resources wireless.
Lappe et al. (2014) have performed a simulation study about such a smart planning system. Their
system calculates the interval between trips for the transport system based on real-time demand
from the production lines rather than using fixed intervals. This has proven to increase reliability
and reduce transportation times by 25 percent. Another proof-of-concept experiment done by
Huang et al., (2008) on smart planning showed that through real-time information traceability they
could make feedback-based decisions for adaptive production control and planning. This improved
decision-making led to increased productivity and product quality. This smart planning first
converts sales orders into the format required for their production system, then plans the total
production of the orders, taken into account constraints like capacity. Finally, the planner can
allocate individual parts to machines, specify the production sequence and then calculate the start
and finish times for all parts. A smart planning allows traditional planning systems to turn into
dynamic production systems through its connectivity, remote adjustability and its ability to
automatically adopt new production schemes (Kolberg & Zühlke, 2015). An example hereof is e-
Kanban, the combination of new technology with the traditional Kanban system (Lage Junior and
Godinho Filho, 2010). The e-Kanban system is more flexible than the traditional system in adapting
to changes, for instance batch sizes (Dickmann, 2007: 405). These smart planning systems do
increase the complexity in operations. For instance, the introduction of a smart planning system as
proposed by Huang et al. (2008) requires smart products, smart machines, smart operators and an
12 ERP system to complete processes. These components store real-time data, which despite its many benefits, increases the amount of data stored and the information available (Schmitt et al., 2017). When this data is not translated into information or used in any other way it becomes waste itself. The handling and storage of all this data further requires an increase in computing power (Luo et al., 2017; Schmitt et al., 2017).
2.3. Process Flow
The process flow performance is often measured by objectives such as throughput rate, cycle time, work-in-progress and resource utilization. These objectives can be used to measure performance and can be used as tools in the identification of waste. Important processes in the process flow are planning and control, capacity planning and control and inventory control.
2.3.1. Planning and Control
We will discuss planning and control both in capacity and inventory context to operationalize the processes where buffer removal is important. The planning operations combine two important processes for a company: planning and control. These processes are used interchangeably in literature (Slack et al., 2010: 270), however the distinction we make between the two is the following: planning is defining what the company wants to happen in the future, control is dealing with unforeseen deviations from the intended path. A wide range of deviations can occur, for example variation in demand (increase in arrival rate) or unforeseen downtime of machines (decrease of capacity). The internal planning and control processes are important to master because of its importance in today’s economy. During the introduction, we stressed the importance of flexibility in answering to ever changing customer demand in a global market. This flexibility means the production system ability in adjusting to customer demand. More specifically, this means planning and controlling in order to meet customer demands of customization, globalization and short delivery times (Harris and Harris, 2008). An important aspect of this is the production system’s capacity planning and control.
2.3.2. Capacity Planning and Control
We will define capacity in operations as “the maximum level of value-added activity over a period of time that the processes can achieve under normal operation conditions” (Slack et al., 2010:
324). Capacity planning and control is then handling the capacity of operations in such a way that it
can respond to the demand placed on it (Harrison & Petty, 2002: 213; Slack et al., 2010: 324). As
defined in the section above Capacity control will focus on the short to medium term, dealing with
unforeseen deviations. Being able to handle these deviations requires a certain amount of
flexibility. Dealing with flexibility in capacity management is where planning becomes important.
13 Because when demand and capacity are precisely aligned, a company might be unable to cope with variability (Slack et al., 2010: 118).
2.3.3. Inventory Control
Inventory control is the operation of controlling inventory anywhere in the operations, from raw materials to finished products. This determines largely the flexibility and speed of an operation. The Toyota Production System became popular because of the way it handled inventory and buffers.
Western firms noticed that inventory control as a process could have significant effects on performance. When companies keep large amounts of inventory and buffers, they often do this to be able to respond to variations in demand (Goldratt, 1990; Johnsen, Howard, Miemczyk, 2014:
281), however inventory and buffers decrease the flow and therefore the flexibility of the company (Goldratt, 1999: 32). This decreased flexibility is partly explained by the increase in cycle time when WIP increases (Little & Graves, 2008) or finished goods increases (Cachon & Olivares, 2010). This then makes customized orders hard to process, as they will take a long time, and may cause a customer to baulk or renege and go to a competitor (Slack et al., 2010: 324). Furthermore, the removal of excess inventory is often seen as the first steps in removing waste as this allows for better identification of problems (Slack et al., 2010: 468). Finally, inventory control can be a costly endeavour, with often-overlooked costs like lost interest on capital stuck in inventory (Karsten, Slikker and van Houtum, 2012).
2.4. Relation
The literature discussed above shows us some mixed results. On the one hand, smart industry implementations can help a company becoming Leaner through reducing errors and failures (Lee et al., 2015; Ohno, 1988), reducing downtime, rework and scrap, improved traceability (Zhong et al., 2013), improved shop-floor management (Qu et al., 2012) and other benefits named above. On the other hand, the implementation of smart components increases complexity as more tools and technology are required to complete the processes, it further increases the sheer amount of information available and computing power required to handle this (Luo et al., 2017; Schmitt et al., 2017). This in turn can lead to information waste and can increase complexity (Zuehlke, 2010).
These tools and technology can also increase the risk of errors and failures. These all contradict
the principles of Lean. As we expect that certain processes are of such a complex nature it is
simply impractical to implement Lean principles, these are the processes where we expect smart
industry components to be able to improve the internal process flow. To investigate the relation
between the two concepts we will make propositions following the literature above. Smart
increases the total data collected, however not all data may be used (Zuehlke, 2010). This creates
proposition 1:
14
“Data collection without creating information leads to information waste”.
Various literature found an improved traceability and shop-floor management (Zhong et al., 2013;
Qu et al., 2012). Because the first steps in removing waste is the identification of waste (Goldratt, 1990: 8) we expect this to allow for Leaner processes. Therefore proposition 2 is:
“The use of smart industry components improves traceability and shop-floor management, these allow for Leaner processes”.
Lee (2015) and Luo et al. (2017) found that smart components improve communication. Literature across industries agrees that communication is vital and can decrease the risk of errors (Slack et al., 2010: 418)
1. From this follows proposition 3:
“Improved communication through Smart industry allows for a reduction in errors”.
Proposition 4A follows from the increase in complexity (Luo et al., 2017; Schmitt et al., 2017;
Zuehlke, 2010):
“The increased complexity caused by smart industry components increases the risk of failure and errors”.
Contradictory, research has indicated that smart components can also have the opposite effect (Lee et al., 2015; Ohno, 1988), leading to proposition 4B:
“The improved traceability and information overview caused by smart industry components decrease the risk of failure and errors”.
Proposition 5 comes from a possible increase in throughput (Orr, 1997), a decrease in downtime (Luo et al., 2017), and a possible optimum between flow and capacity enabled by smart planning (Kolberg & Zühlke, 2015). These allow for a better response to demand placed upon the capacity of operations, leading to proposition 5:
“Smart Industry components improve capacity management”.
3. RESEARCH METHOD 3.1. Research Design
In order to create a better understanding of the relationship between smart industry, the internal process flow and Lean, a single-case study is performed. The case we selected has implemented a smart Industry system in the internal logistics. This allowed us to study the effects of smart Components on a part of the process flow. The company is also looking to implement a new smart Industry project in different processes. This will allow us to both investigate the previous
1For literature about communication and errors in different industries we recommend: Helmreich, R. L.
(2000). On error management: lessons from aviation. BMJ: British Medical Journal, 320(7237), 781.
Frydenberg, K., & Brekke, M. (2012). Poor communication on patients’ medication across health care levels leads to potentially harmful medication errors. Scandinavian journal of primary health care, 30(4), 234-240.
Sutcliffe, K. M., Lewton, E., & Rosenthal, M. M. (2004). Communication failures: an insidious contributor to medical mishaps. Academic Medicine, 79(2), 186-194.
15 implementation and its results, as well as investigate the possibilities of a new implementation.
Our research will investigate the propositions we developed following contemporary literature. The keywords used to find literature are provided in Appendix C. We made use of ERP-data analysis, interviews and observations to investigate our propositions. These preliminary results are discussed through in-depth interviews. Because we want to find out how smart industry developments can improve the process flow whilst following Lean principles and because relations between these variables are not quite clear, a case study is the proper research method (Eisenhardt, 1989: 536; Voss, Tsikriktsis, & Frohlich, 2002; Yin, 2011). This allows for a more detailed, contextual analysis of the processes and events (Blumberg, Cooper & Schindler, 2008).
Our focus will lie on the operational level of the organisation (Figure 2) and the levels of planning involved. The operational level also includes the operational (process) control.
Figure 2. The levels of decision-making and their corresponding planning system (Harrison &
Petty, 2002: 8).
Recent literature has discussed what smart Industry entails conceptually, however there has been
little to no empirical research performed (Yoon et al., 2012). This becomes most apparent when
researching smart Industry and the planning processes, in which multiple authors use the term
smart Planning but with varying meanings (Appendix B). This is because of a lacking general
definition. This case study hopes to add to the currently scarce amount of empirical research
available in the field. The case that is studied will be referred to as Company X. Company X was
selected based on several criteria which we will elaborate upon in the next section.
16 3.2. Case Introduction
The case selection was based on two primary criteria. The company had to be Lean to a certain degree or trying to become Lean. Secondly, the company had to be interested in smart industry and preferably it should already have implemented components of smart industry. We selected these criteria because of the explorative nature of our research; we require a case study with interest and knowledge about the subjects for our research. This will ensure data collection with people who are familiar with the subjects (and in turn prevents misunderstanding). We reason that this will also increase commitment to our research, as people who have no interest in smart Industry are less likely to support research about it. We assessed the degree of smart Industry integration at Company X using the model created by Lee (2015) (Appendix A). Using this model, we find that Company X is in the ‘Smart Connection Phase’ (level 1) with the most important indicators being accurate and reliable data from IoT-based machines and sensors. Furthermore, the presence of ERP and wireless communication are strong indicators. Levels above 1 are not achieved as Company X lacks further integration of smart components and machines. The assessment of our selection criteria increases the replicability of our research and increases the external validity (Yin, 2009: 226). Company X is a large manufacturer (>500FTE’s) in the chemical industry and has 7 plants spread over the Pearl River Delta. Our research will focus on one of these plants. Company X works primarily on a Make-To-Order (MTO) basis (Figure 3). This is because their products are not produced until the order has been confirmed (Rudberg and Wikner, 2004). The fact that Company X works on a MTO basis has significant implications for its process flow. MTO production style requires flexibility in the production and warehouse management, and more importantly, requires a focus on short throughput times as Company X will have to be able to fulfil the orders within the promised delivery time (Jammernegg & Reiner, 2007).
Figure 3. Typical CODP (Olhager, 2012)
22 Research has indicated that grouping the CODP in four categories is oversimplified. However, this is outside of the scope of this research (For more information on this subject we recommend: Rudberg &
Wikner, 2005).
17 Company X’s processes are simplified in figure 4. The current average throughput time is 72 hours of which the internal logistics and warehousing takes an average of 29 hours. The currently implemented IoT system involves transporting finished goods after production into the finished goods warehouse (Figure 4).
Figure 4. Simplified model of the process flow and IoT system at Company X.
The internal logistics and warehousing processes, from the end of production until external logistics, is our unit of analysis. These are the processes that are smart and using these processes as our Unit of Analysis will allow us to investigate the relation between Lean, smart and the process flow. The analysis of these variables, but mostly flow, requires an integral view of the processes involving internal logistics and warehousing. Company X has 3 finished goods warehouses, 2 of which are directly connected to external transport. A simplified map of the whole plant can be found in appendix D.
3.3. Data Collection
The data was collected using a single case study with multiple sources of evidence: open face-to- face interviews, observations and analysis of secondary quantitative data. This allows a full grasp on the variables at play. Furthermore, the multiple sources of evidence allow us to triangulate our results (Yin, 2011: 8). The triangulation will increase the construct validity and reliability of the research (O'Donoghue & Punch, 2003: 78; Maxwell, 2009: 244-245; Voss et al., 2002). The research will be performed by means of an internship, allowing for an in-depth understanding of situations and the opportunity to make repeated observations and interviews. This will increase the validity of the research (Maxwell, 2009: 244-245). We have created propositions following the literature about the variables in our research question. We will analyse these propositions using our case study and by doing so we hope to find an answer to our research question. The research was performed in three phases, starting with the exploratory phase, this phase is meant to get the overview of company, its employees, the production processes and the current planning system.
We focused on the current planning processes, the current level of Lean and already implemented
smart Industry components. This phase allowed us to find the relevant processes and people from
which to collect data. The second phase focused on collecting and analysing data about the IoT
18 system, buffers, variabilities and preliminary thoughts about our propositions. Multiple guided tours amongst workstations, warehouses and other processes onsite were performed. Additionally, face- to-face interviews were performed. All of the guides and interviewed people are shown in the organigram below (figure 5). The process of interviewing started with collecting data during a guided tour along a process, then discussing and asking open questions during the tour. During the tour, field notes were created of observations and discussions. These helped in identifying waste, variability and buffers. Using the data collected during the guided tour we created a semi- structured interview guide. We would then continue with semi-structured interviews, which took between 1 to 4 hours. The interview and observation guide can be found in Appendix E.
Performing the interviews after the guided tours allowed us to gain in-depth information from insights that arose from the guided tours, and it allowed us to confirm observations.
Figure 5. Organigram of tour guides and persons interviewed at Company X.
Furthermore, in order to further review our propositions, we analysed ERP- and hand-collected
data. To do this we used tools such as throughput diagrams and spreadsheet analysis. Through
the combination of observations, data analysis and interviews we were able to analyse the flow
and diagnose various sources of waste. The final step was discussing our results through open in-
depth interviews with top management and the process managers. This allowed us to obtain
feedback on the collected data in order to improve the respondent validity (Maxwell, 2009: 244-
245).
19 3.4. Design
We have used the findings of the internal logistics system in our case company to design a new smart project for the remainder of process flow, from raw materials until the end of production (where the current smart system starts). This involves 9 active workstations, 2 finished goods warehouses and 4 raw material warehouses (a simplified map of the entire plant is attached in Appendix D). In order to analyse the process flow we started with observations and short interviews, followed by in-depth interviews. We combined these with analysis of secondary quantitative data. We used throughput diagrams, spreadsheet analysis and the calculation of the synchronization time. These methods allowed us to find various sources of waste and find several other potential improvements in the process flow. Using our findings of the internal logistics case we have designed smart solutions to these potential improvements.
4. RESULTS
The results are constructed as follows: first, we will present the IoT system in the internal warehousing and logistics at Company X. We will then review the propositions we made in the literature review. Afterwards, we will discuss our results.
4.1. Internal Warehousing and Logistics
The IoT system at internal logistics and warehousing contains multiple smart Industry components.
These components and their information flows are demonstrated in figure 6. The case company
makes use of barcode technology on the products and QR technology on the pallets. These
contain information such as customer information, product information, deadlines or order
information. Furthermore, the operators make use of mobile computers and tablets. These allow
them to access information on the products at any time and make adjustments to the planning,
forklifts are equipped with scanners and screens that are both connected to the internet. The use
of smart products, smart operators and smart machines enables the case company to use a smart
planning system. Smart machines and operators scan smart objects and by doing so transfer data
wireless to the smart planning system. The smart planning uses the ERP system to register
information related to the smart objects. The smart planning uses this information along with ERP
data to decide which items are due for pickup and their target location in the warehouse (logistics
and warehouse scheduling), and transmits this information back to the smart operator and/or smart
machine. An optimal route is created through optimization using three variables: the location of the
items, the operator/machine and finally the target location. Smart operators can manually override
the logistics and warehouse planning in order to make room for emergency orders or other
priorities. The IoT system uses wireless real-time data, translates it to information and operation
20 resources can be adjusted wireless. According to our definition the IoT in the internal logistics and warehousing can be defined as a smart planning system.
Figure 6. Information flows at the IoT system in warehousing and internal logistics.
The IoT system starts when production is finished and employees at the workstation perform a quantity check and combine a pallet (QR code) with a product (barcode) using a mobile computer.
From that point, the smart planning schedules a forklift driver to pick up the products. The forklift
reader informs forklift drivers which products to pick up (objective) and the target destination. This
eliminates waiting time for forklift drivers as the next objective is shown on the screens inside the
forklift (top left in figure 7). Furthermore, this also prevents transporting the wrong items as the
forklift scans the QR code on the pallets to confirm the pickup is the same as the objective. This
simplifies the forklift drivers work and decreases variation as searching is practically eliminated and
nearly no communication is required between employees. The smart objects contain all information
that is necessary to complete the task. Warehouse employees can access the same information
using tablets or scan smart objects with mobile computers.
21 Figure 7. Simplified depiction of the IoT system in practice.
4.2. Analysis of the current IoT system
The propositions we made at the end of our literature will be reviewed using the IoT system. The first proposition was: “Data collection without creating information leads to information waste”. To review this proposition we analysed the database, containing all real-time data, and we observed what information is shown on the smart components. During observations, we found that only the information necessary was shown on the appropriate smart component. For instance, the data shown on the smart forklifts is far more limited than those shown on the shop floor tablets. An interview with the person responsible for the IoT implementation confirmed that this was of high importance during implementation. Because of the overwhelming data available or the complexity thereof it was deemed necessary to create automated views for different people and components.
A good example of this is the data the forklift driver has access to (bottom and mid left in figure 7).
He can only see his objectives and target locations. In turn, there is also limited data collection as he can only enter into the system whether he has completed his task and the quantity transported.
Information like due dates are not relevant to the forklift driver and thus not shown. However, after analysing the ERP database still a large part of the data collected is not being used and this data does get stored, leading to information waste. The IT manager told us that originally extra data was collected for possible improvements. So far, this data has gone unused.
The second proposition was: “The use of smart industry components improves traceability and shop-floor management, these allow for Leaner processes”. We found that the smart components implemented at the internal logistics and warehousing improved traceability. As we observed, the smart components provided employees with current locations of goods and with upcoming tasks.
During a guided tour, we found that this was a significant upgrade for forklift drivers. They no
22 longer had to wait for information on which products to pick or where to place them. As expected it also virtually eliminated the need for searching for items. The production manager and workstation managers stressed the importance of the increased traceability, explaining that the success of the IoT project was decided by its ability to create an overview of “the real-time execution status, such as the occupancy of all the locations in the warehouse, in each stage of the warehouse processes”. The smart components greatly improved the shop-floor management, according to management. The improvement in shop-floor management can be explained through two features of the smart planning. First of all, the planning system already creates an optimized schedule and thus greatly simplified the shop-floor management task. Secondly, it allowed management on the shop-floor to adjust priorities and thus dynamically adjust the order of shipping as necessary. This reduced waiting time (both of items and employees), unnecessary transport (or wrong transport), and the more efficient way of shop-floor management lead to a reduction in WIP in the finished goods warehouse (reduction of inventory). The improved shop-floor management also reduced variability in the warehouse. The smart planning schedules the arrival of finished goods by the forklifts across the lanes and warehouses based on their shipment due date. This minimizes variability in arrival rate for the warehouses. The introduction of smart however did cause technical issues with the connection between smart components and the system. We will go further in detail about this in proposition 4A.
Our third proposition is: “Improved communication through smart industry allows for a reduction in errors”. We found support for this proposition. Smart components improved communications and thereby reduced errors in two ways. First of all, through short interviews on the shop floor we found that direct communication between actors required for operations was greatly reduced because most information is communicated through smart components. This reduced miscommunications as direct communication between for instance forklift drivers and shop-floor management is no longer required. Smart components prevent misunderstandings as smart components allow for a higher complexity of information to be transmitted to each other. Forklift drivers stressed that information on a screen is more obvious than receiving information vocally or by telephone in noise environments such as a logistics warehouse.
We will discuss Proposition 4A: “The increased complexity caused by smart industry components
increases the risk of failure and errors”. We found support for proposition 4A. The complexity of the
processes has increased through the required constant connectivity between the components and
the system. This constant internet connection is necessary in order for the smart components to
exchange information with the smart planning, for example a forklift to receive his next pickup
location or a shop floor manager to scan a QR code. However, during observations we noticed that
in certain locations of the warehouses the signal was too weak, causing connection losses for the
smart components. This caused shop-floor operators and forklift drivers to exit the shop-floor and
23 wait for reconnection in order to enter data into the system. In this case the required connection between the parts increased complexity which has led to failures: the failing connection between the components and the system. This caused errors in the data such as incorrect timestamps and missing data. We therefore have strong support for proposition 4A. At the same time, we also found support for Proposition 4B: “The improved traceability and information overview caused by Smart Industry components decrease the risk of failure and errors”. During interviews with the shop-floor managers and production managers we found that the increased traceability and better overview of information has led to a decreased amount of errors, as mistakes in placement of items are no longer made. There is no confusion about where products are stored and the system reduces errors such as products in wrong locations or lanes. Furthermore, loading errors are significantly reduced. Management indicated that in the past mistakes were sometimes made because of missing papers on pallets (most of Company X’s products are packaged in the same material and the only identification was a product information sheet that was easily lost). Through the use of barcodes and QR this is no longer a problem.
Finally, our fifth Proposition was: “Smart Industry components improve capacity management”. We found strong support for this proposition. The primary resources in logistics are the forklifts and the finished goods lanes (Figure 7). Management and Forklift drivers indicated that they have a substantial reduction in waiting time (for assignments or objectives) and unnecessary motion (searching). The system further creates optimal routes depending on the forklifts location, the items’ location and the target location. Performance review of the IoT system used by management analysed a month of data from both before and after IoT, this showed that the average transportation time has been reduced with 38%. The finished goods lanes are where items are placed for final transport to the customer. The smart planning optimizes these lanes based on due dates, capacity and flow. The same performance review used for the transportation time also compared a month before and after IoT on the finished goods lanes. This review showed us that the average time items spent in the lanes has been reduced with 19%. These improvements have increased the level of value-added activity under normal operating conditions, in other words increased capacity. Additionally, the automatic optimization of the lanes enables management to only manage the capacity of the lanes when there are irregularities such as emergency orders, simplifying their work.
4.3. Discussion
In this section, we will start with a brief discussion of the direct effects of Smart on the process
flow. This is followed by discussing the paradox of smart versus Lean that started this research.
24 4.3.1. Smart and Process Flow
We found various other ways that Smart components influence the process flow that are not directly defined under the Lean improvement approach. Smart Industry can play a great role in capacity management as it decreases waiting and transportation times and allows for a more efficient way of working. However, we found that when there are no capacity issues smart industry components find themselves more in a supporting role. It is comparable to creating a poke-yoke environment, a mechanism in Lean that helps employees in avoiding mistakes. Smart industry components simplify work and prevent errors. It also simplifies communication and prevents ambiguity concerning objectives.
We expected there to be a connection between downtime of resources (e.g. for maintenance) and smart as recent research had found (Luo et al., 2017). However, our case company did not make use of predictive analysis to predict failures or downtime of resources and thus there was no connection between the two. When discussing this they clarified that for predictive maintenance of resources they did not require a smart system as this is currently planned on usage and this works satisfactory. However, for resources where required maintenance and wear is unpredictable we argue that smart is a good solution to indicate maintenance. As Yoon et al. (2012) describes in his conceptual framework for the smart factory, the use of sensors can recognize potential resource risks and based on the context respond accordingly.
Finally, we found information waste at our case company. However, our case company indicated that information waste was relatively unproblematic. Storage of (cloud) data is cheap and the use of cloud storage rather than local storage allows for reliable data storage. It must be noted however that our case company has not used the extra data for any potential improvements. We argue that when improvements are sought, the extra data will also cause unnecessary work as employees will have to search for the right data.
4.3.2. Lean versus smart
The paradox from our introduction becomes less apparent in our case study. We expected that because of the paradox between Lean and smart, the combination of the two would cause issues.
What we actually found is that our case company is using IoT to achieve a Lean and that in fact for the greater part a synergy between the two exists. As one respondent perfectly summarized: “Lean and IoT are two different things, we use IoT technology to help us achieve Lean. We use IoT to get the real-time data, this can be used for improvements”. Some authors presented smart as a whole new paradigm (Yoon et al., 2012; Lucke et al., 2008; Meyer et al., 2009; Zuehlke, 2010; Zuehlke, 2009), while others saw it as an improvement on current methods (Kolberg et al., 2016; Kolberg &
Zühlke, 2015). We agree with the latter; in our case company Smart was used as a tool to facilitate
25 Lean. The relationship between smart components, Lean and the internal process flow as we found it in company X is demonstrated in a conceptual model (Figure 8). The direct positive effect smart has on the process flow is explained in the section above (4.3.1).
Figure 8. Conceptual Model
Smart components have this mixed effect on Lean through the multiple ways that smart components influence waste. The smart components have simplified operations for employees. As we already found during our literature review, this was expected to be a great source of reducing waste (Slack et al, 2015: 472). There were however exceptions. As Lee et al. (2015) and Ohno (1988) mentioned, smart industry may cause errors, which is indeed what we found. We have summarized the various waste types that were influenced, either positively or negatively, by smart components in table 1. When our findings were in line with literature we mention the references after the ‘How’.
Waste How Smart Components
involved Waiting Time + Information (e.g. objectives and
locations) on smart screens eliminates waiting time of forklift drivers and location resources.
- Dropped connections sometimes forced shop floor workers and forklift drivers to wait until reconnection with the system.
• Smart Forklifts
• Smart Tablets
Transport + Placing the finished goods directly in the final lane for pick up eliminates the possibility of moving items around unnecessarily.
+ Dynamic routing scheduling of forklifts minimizes transport distances (Kolberg
& Zühlke, 2015; Lappe et al., 2014; Qu
• Smart Forklifts
• Smart Planning
• Smart Pallets
26 et al., 2015; Wang et al., 2015; Xu,
2011).
Inventory + Increase in transparency and traceability has led to a reduction of the time finished goods spends in the warehouse, leading to a total decrease of inventory (Qu et al., 2012).
• Smart Pallets
• Smart Objects
• Mobile Computers
• Smart Tablets
• Smart Forklifts Motion + The simplification of the operations on
the shop floor prevents superfluous and unnecessary motion of employees.
+ Increased traceability and overview prevents unnecessary searching for items (Qu et al., 2012).
- Dropped connections sometimes forced shop floor workers and forklift drivers to move in order to reconnect to the system.
• Smart Tablets
• Smart Forklifts
• Mobile Computer
Defectives + The use of QR technology on pallets prevents transporting wrong products (Huang et al., 2008; Zhong et al., 2013).
- Dropped connections caused errors in the database (Lee et al., 2015; Ohno 1988)
• Smart Objects
• Smart Forklifts
• Mobile Computers
Over-production - The storage of unused data has led to an overproduction of data (Zuehlke, 2010).
• Smart Planning
Table 1.
5. Design
In this section, we will first discuss the process flow to identify any sources of waste, learning more
about the relation between Lean and the process flow. Using these sources of waste and our
findings on how Smart influences waste and the process flow we create a smart Design to remove
the identified waste sources. To create this design, we will use our literature review and previous
findings. The identification of waste in the process flow and then designing a solution for this waste
has led us to identify new ways that smart, Lean and the process flow influence each other.
27 5.1. Process Flow and Problems relating to Smart
We will now discuss the process flow from the moment of selecting the raw materials to placing them outside the workstations (figure 9). Each of the workstations goes through these processes.
5.1.1. Processes
The process starts with selecting raw materials from either the warehouse or from buffers inside the workstation. These raw materials are placed inside mixers. The process of filling the mixers differs on the order size as certain sizes of mixers are in different locations and the process of filling them varies. The mixing is followed by a quality control. After the quality control (meaning no rework is required) the paint gets distributed among individual containers. The filling process involves placing bins on an elevator (called a filling spot) which allows employees to place individual containers beneath the mixing bins. After the individual containers (containing a barcode) are filled they are put outside the workstation on a pallet (containing a QR code). Here the containers wait for transportation into the finished goods warehouse.
Figure 9. The process flow inside a workstation.
5.1.2. Capacity
The capacity of workstations depends on the amount of people working, mixing bins available and
filling spots. We found that there are generally enough people working, if they suspect any capacity
shortage they add temporary staff. Workstation managers indicated that this decision is based on
experience. The mixing bins and filling spots always have enough capacity according to the shop-
floor managers and observations. Observations and interviews indicated that the workstations
have a surplus amount of capacity, even during peak times they can still handle demand. However,
from our literature review and our previous findings, we found that smart industry can positively
influence the capacity of resources. Therefore, we wanted to verify they did not have a capacity
problem in the workstations. We analysed a month of throughput data from every workstation. A
throughput diagram for one of these workstations can be found in figure 10. Time is on the X-axis
(around 2 months of data) and the cumulative workload on the Y-axis. The workload is the amount
of work (in minutes) an order will take. The throughput diagram shows us the relation between the
confirmed amount of work, the planned work, the completed work and the due dates. These
showed us that Company X is under-loading the workstations. Even during an increase of work
28 (planned input) the completed output increases at the same degree, indicating that there is enough capacity to handle variability.
Figure 10. Throughput diagram of a workstation.
5.1.3. Quality control
The quality control is performed after mixing. Should an item not pass this quality check it has to be reworked. This means going back to the mixing station and having ingredients added. This will be repeated until the product passes the quality check. Interviews with shop floor employees indicated that generally only 60% of the mixed bins pass this check, making it a large source of defectives waste. This is caused by 2 issues. First of all, due to the raw materials not being up to quality standard. Most products are sourced from multiple suppliers and these suppliers vary in product quality. By sourcing their raw materials from multiple suppliers, they now have to keep track of thousands of different raw materials in varying quality. The raw material manager confirmed that the large number of different products makes it hard to keep track of differences between individual suppliers. The second problem is employees adding the wrong quantities of raw material. The mixing instruction per product type are on printed paper and these make no difference between raw materials from different suppliers (say product A from supplier B or product A from supplier C).
5.1.4. Raw Material Management and Starvation
The average throughput time of an order is 3 days. In order to achieve this there cannot be a stock-out of a certain product. Ordering is based on a reorder level combined with a naïve forecast system (reordering last year’s usage). To prevent stock outs the warehouses have stored large amounts of raw material. Material usage is tracked using the ERP system, however this turned out to be unreliable. Therefore, the warehouse manager has to perform a daily check of current stock quantities. The unreliable tracking of material usage by the ERP system is caused by variations in
0 1000 2000 3000 4000 5000 6000 7000
24-Feb 6-Mar 16-Mar 26-Mar 5-Apr 15-Apr
cummulative workload (min)
time (days)
confirmation planned_input completed due_date
29 raw material and inaccurate or incorrect data entry. Furthermore, after observations and reviewing 3 months of raw material data, we found out that order frequencies and usage of material are completely separated. To demonstrate this, we analysed the amount of specific item types that got supplied without having any withdrawals and items that got withdrawn but never supplied. This can be seen in figure 11. For example, in warehouse 2 there are 27 item types that got replenished when they were not used and 84 that were withdrawn but never supplied. With this system, the naïve forecast method causes stock that is hardly used to build up and at the same time material that is required to be out of stock. During an interview, the raw material manager confirmed this and told us the large amount of different unit sizes and suppliers cause a chaotic situation in the raw material warehouse that is hard to oversee. They got used to storing huge amounts of raw material in order to overcome variations in demand. While starvation is not common at the moment, they only avoid this by keeping huge amounts of stock.
Figure 11. Warehouse withdrawals and supplies
5.1.5. Synchronization
A complication we found after analysing production data using spreadsheet analysis was the synchronization of suborders. Each order is split in multiple suborders, this is because most orders consist out of multiple product types. These suborders are distributed among the workstations. The manager of each workstation then decides the production date of these based on the shipping date and the production time. However, these suborders are not synchronized between the workstations (parallel). This results in suborders waiting for the completion of other suborders. This is what we call synchronization time. We analysed one month of data to find out what the distribution of this synchronization time was between the orders. These results can be found in figure 12
*. These results show us that there is a relatively high synchronization time compared to the average
*This table contains only orders that contain multiple suborders. Meaning that the orders in figure 12 are about 60% of the total orders in the analysed month. The remaining 40% of orders consist out of a single suborder and thus have a synchronization time of 0.
warehouse
Withdrawals but no Supply
Supply but no Withdrawals
0 9 16
1 1 0
2 84 27
3 51 11
5 6 1
6 13 5
7 16 21
8 14 1
2 0
0 0
