Improving Workload Control applicability: mechanisms to address real-life complexities

(1)

Improving Workload Control

applicability:

mechanisms to address real-life

complexities

Master thesis

Author: J.M. van Noort Student Number: S2960303

E-mail address: j.m.van.noort@student.rug.nl Supervisor: dr. J.A.C. Bokhorst

Co-assessor: dr. ir. D.J. van der Zee

MSc. Technology and Operations Management Faculty of Economics and Business

University of Groningen Academic year 2019-2020

Abstract: To comply with the market demand to deliver customized products fast and reliable, many companies are

seeking to improve their delivery performance. This market is largely dominated by small and medium enterprises (SME), that manufacture a high variety of customized products in relative low volumes (HVLV). The metal sheet processing industry is an example of a market with typical HVLV manufacturers. Within the industry of sheet metal processing, it is common practice to “nest” different orders to realize raw material efficiency. Also, these companies often suffer from labour constraints as a large part of operations is performed by operators. Such situation is referred to as dual resource constrained (DRC), and together with nesting, often neglected in Workload Control literature. The complexities of nesting and DRC are in fact limiting the functions of WLC. WLC is a production planning and control (PPC) approach that is considered to be one of the best solutions for improving delivery performance in HVLV companies. That being at least theoretically because implementation of WLC in practice often results in disappointment. This research examines a case company within the sheet metal processing industry for the applicability of the WLC approach. Whilst theory suggests that WLC is of good fit to the company, an analysis on the practical complexities and issues shows otherwise. This research provides researchers and practitioners useful insights on the mechanisms of a WLC based system that considers these complexities.

Keywords: Workload Control, Production planning and control, Make-to-order, High variety low volume, Design

Science Research, Nesting, Dual Resource Constrained, Delivery Performance, Material Efficiency

(2)

Abbreviations

DD Delivery Date

DP Delivery-time Promising Process DRC Dual Resource Constrained DSR Design Science Research FTE Full Time Employee

HVLV High Variety - Low Volume OEM Original Equipment Manufacturer OM Operations Management

PPC Product Planning and Control RBC Repeat Business Customizer RP Realisation Process

SME Small and Medium Enterprises VMC Versatile Manufacturing Company WIP Work in Progress

(4)

Preface

I would like to thank Jos Bokhorst for his everlasting support during the writing of my thesis. He remained confident that I would complete this thesis successfully, even though my efforts on the research proposal were quite poor due to the birth of my son Evan in that period. Jos motivated me to continue with the project, which I greatly appreciate. As it was not difficult enough to play the role of father and researcher at the same time, the Corona-virus outbreak also put extra stress and pressure on everybody involved. I was very pleased to see that everyone switched to online meetings and that the project could continue.

I am very grateful to the case company, who made it possible for me to execute the research process. I have learned a lot during the process and I hope my efforts are of value to the company. Also, Jannes Slomp deserves to be put in the limelight. He was very involved in the project and I could always count on his feedback. Also special thanks to Gerlinde Oversluizen, who helped establish contacts with other students and Tom Wienholts, who provided me with a lot of little bits of missing information.

(5)

1. Introduction

High variety, low volume (HVLV) manufacturing concerns the manufacturing of a variety of customised products in relative low quantities and is mostly common with small and medium enterprises (SME’s) following a make-to-order (MTO) strategy (Amaro, Hendry, & Kingsman, 1999). HVLV is considered to be one of the most challenging manufacturing strategies to control, as the ability to produce a wide variety of products also tends to introduce great variability in the system (Gran & Alfnes, 2019; Katic & Agarwal, 2018). SME’s that operate within an HVLV setting often wish to improve their delivery performance (M. J. Land, Stevenson, Thürer, & Gaalman, 2015; Missbauer & Uzsoy, 2020; Soepenberg, Land, & Gaalman, 2012a). Work load control (WLC) is considered to be one of the best product planning and control (PPC) solutions to improve delivery performance in a HVLV environment (Cransberg, Land, Hicks, & Stevenson, 2016; Hendry et al., 2013; Slomp, Bokhorst, & Germs, 2009; Soepenberg, Land, & Gaalman, 2012a; Stevenson et al., 2005; Thürer et al., 2012). WLC is a robust PPC solution that releases orders to the shop floor to balance workloads, whilst also buffering against the uncertainties of incoming orders (Stevenson et al., 2005). See Figure 1 for a visual representation of the WLC approach. The approach has gained much attention because of its simplicity and (theoretical) effectiveness. However, it has shown that ever since the emergence of WLC, theoretical attempts are often distant from practice (Hendry, Huang, & Stevenson, 2013; Missbauer & Uzsoy, 2020; Thürer, Stevenson, Silva, & Huang, 2012). A gap between practice and theory is unfortunately common in the field of Operations Management (Holmström, Ketokivi, & Hameri, 2009). For WLC, the gap between practice and theory is partly a result of complexities that make it difficult to implement WLC in practice (Cransberg, Land, Hicks, & Stevenson, 2016).

Figure 1: The WLC approach (Mj Land, 2004)

(6)

Adapting a HVLV manufacturing strategy creates many uncertainties and complexities (Katic & Agarwal, 2018). Within the metal sheet processing industry, one common observed complexity is the nesting of orders (Cransberg et al., 2016). With nesting, several orders are allocated to sheet(s) of raw-material for cost-efficiency purposes (Herrmann & Delalio, 2001). Nesting tends to add great variability to the system, as a group of orders with different delivery dates is simultaneously released to the shop-floor. Furthermore, the aim for cost-efficiency at the process of nesting conflicts with making optimal load balancing and timing decisions at the release function of workload control (Cransberg et al., 2016; Mj Land, 2004). Another complexity that is often not considered in the literature on WLC, is the fact that these companies may also be constrained by not only machine capacity, but also by labour capacity (Thürer, Stevenson, & Renna, 2019). This is referred to as a dual resource constrained system (DRC) (Bokhorst, Slomp, & Gaalman, 2004).

To date, PPC literature has not yet tried to integrate the complexities of nesting and DRC into the concept of WLC (Cransberg et al., 2016; Mark Stevenson, Huang, Hendry, & Soepenberg, 2011; Thürer, Stevenson, & Renna, 2019), which seems to conflict with the advocated appropriateness of the WLC approach in a HVLV setting. Luckily, the rise of modern technologies creates new possibilities for advances in the area of PPC (Missbauer & Uzsoy, 2020). The objectives of this study are to narrow the gap between theory and practice on WLC. This is realized by identifying complexities that are constraining the functions of WLC and a thorough discussion of these complexities. A solution is proposed, based on WLC, that deals with the complexities. In order to do so, this research follows a Design Science Research (DSR) approach. DSR foresees the need for more practical relevance, as DSR seeks to develop knowledge by engaging in real-life situations (van Aken, Chandrasekaran, & Halman, 2016). The solution is developed following an iterative process, using a combination of case-based data, qualitative information and a knowledge base of existing solutions, systems and theories. To prove relevance and validity, the design will be validated by the case-company.

(7)

2. Research objectives and methodology

2.1 Background

The appropriateness of WLC at the case company is assessed using the framework of Henrich, Land, & Gaalman (2004), with results being promising (see appendix: Assessment for WLC appropriateness). From a literature point of view, it seems that a WLC approach is of good fit to the company. However, the practical applicability of WLC has been questioned in literature by many, which is known as the

WLC paradox (M. Stevenson, Hendry, & Kingsman, 2005). Its practical applicability has never seemed

to meet the promises of its theoretical promotors. For the past few years, literature has started focussing on narrowing the gap between theory and practice by identifying possible factors that hinder the mechanisms of WLC (e.g. Cransberg et al., 2016; Fernandes, Land, & Carmo-Silva, 2016; Gran & Alfnes, 2019; Soepenberg, Land, & Gaalman, 2012c). This study contributes to this literature stream by trying to identify and elaborate on complexities that hinder the application of WLC in its current form and generate a solution that can provide both researchers and practitioners with useful insights.

In line with the wish of managers of SME’s, who are generally not looking for very expensive and time-consuming solutions (de Man & Mannhardt, 2019), it is argued that mechanisms of the “proven” WLC approach should be retained, rather than developing a complete new PPC system. WLC has some very potent, yet simple mechanisms to control a complex system. However, in its current form, WLC does not account for complexities in practice.

2.2 Research objectives

A PPC system that functions following the core mechanisms of WLC while also satisfying complexity aspects, has the potential for great performance improvements in practice. The design of mechanisms that connect WLC with complexities, is therefore the core element of this study. In order to do so, the objectives are to: First identify the PPC related complexities in practice. Secondly, perform an in-depth analysis on how these complexities are constraining the functions of WLC. Thirdly, design a solution that deals with these complexities, integrating them in the WLC principles. Fourth, discuss how these newly developed mechanisms contribute to improving delivery performance in practice. And lastly, generalize the solution in order to inform other researchers and practitioners in the field of Operations Management of the findings of this study.

2.3 Methodology

Transparency is key in research. This section will elaborate on the methodological aspects of this study. It should provide thorough understanding of the research process that is conducted. This study positions itself within the Operations Management (OM) discipline. OM is very practical in nature, which often causes theory and practice to be out of sync (Liu, Vengayil, Zhong, & Xu, 2018; van Aken, Chandrasekaran, & Halman, 2016). This is supported by recent suggestions of Cransberg et al. (2016); Fernandes et al. (2016); Missbauer & Uzsoy (2020); Thürer, Stevenson, Land, & Fredendall (2019) to include more empirical evidence from specific cases in practice. To comply with this desire, this study will follow a Design Science Research (DSR) strategy. As DSR is very interactive with the real world, it is no surprise that this research method has found its way to OM (van Aken et al., 2016). Previous OM research has largely been based on the principle of hypothesis formulating and then testing these hypotheses. However, that strategy is becoming somewhat outdated, as it does not allow for reaping the benefits of solutions that arise from practice (Holmström, Ketokivi, & Hameri, 2009).

(8)

a generic design and at the same time also values the context-specific outcomes. “DSR is a science of

the average (e.g. by developing a certain type of bridge) as well as a science of the particular, giving knowledge on how to deal with specific contextual issues (such as in designing an instantiation of a bridge over a river with unstable shores)” (van Aken et al., 2016).

“A DSR project typically is driven by a type of field problem ... or an opportunity” (van Aken et al., 2016). Both a field problem and opportunity co-exist in the motivation of this study. The case company used for this research is an example of a company that experiences a field problem, being the poor delivery performance. Opportunities arise from new technologies, such as an algorithm or system that foresees in decision making. The field-problem may be solved using these opportunities risen from new technologies.

A DSR study usually consists of: (a) an explanatory component, (b) a design component and (c) a knowledge base (Hevner Alan, 2007; van Aken et al., 2016). The connection between the first and the second component is often referred to as the relevance cycle. In this cycle, the goal is to assure relevance between the design and the real-life situation it concerns. The connection between the design and the knowledge base is referred to as the reflective or rigor cycle. This cycle aims to translate the design and learning activities to generic academic contributions to the field of OM (Hevner Alan, 2007; Wieringa, 2014). These components are interrelated and should be present in every DSR project. Furthermore, two criteria that are of key to DSR in OM: (1) the pragmatic validity (does the study produce the desired outcomes?) and (2) the practical relevance (exploiting an opportunity or a valuable contribution to a significant field-problem). In Figure 2, the research methodology of this study is visualized.

Figure 2: Adapted from Hevner (2007) and van Aken et al. (2016)

(9)

contributes to the validity and relevance of this research. Lastly, it must be noted that DSR is no straightforward procedure of research; it can be operationalized in various ways (van Aken et al., 2016). For the positioning of this research, it is important to gain in-depth understanding of the case company. This enables future researchers and practitioners to relate to these specific characteristics. Therefore, the following section (3) will include a detailed description of the current PPC related aspects of the current system. Here, the root-cause of the poor delivery performance is traced, which will further shape the focus of this research. Other complexities and resulting issues are also discussed in this section. Section 4 includes the theoretical background on complexities in relation to mechanisms of WLC as well as a discussion of recent literature in the field. In section 5, the proposed system is explained and in section 6, the system is validated by feedback of the case company. The paper closes with a conclusion of the findings and recommendations for future research. The structure of the research is visualized in Figure 3.

(10)

3. Current System

This section will first provide some general information with regard to the case company. The current product planning and control system is discussed in the light of WLC control points. The section closes with an overview of complexities and issues that play an important role in the company. This section also includes a detailed analysis of the current delivery performance, which results in a root cause that will be of main interest in the design of a solution.

3.1 General information

The case company manufactures various “ready-to-install” products for OEM’s. The company is able to manufacture virtually anything that is based of sheet metal. Typical products include prefabricated metal profiles for the construction of storage racks, metal body work for touring-busses and heavy-duty safes. Operations include (laser)cutting, folding, pressing, punching, welding and assembly.

The company is currently employing around 45 full time employees (FTE), has an estimated turnover of €5 million and can therefore be classified as an SME. Most products are made following a make-to-order (MTO) principle. Products are mostly made as specified by the customer and this results in a great variety of products that are produced. The orders can be of any size, from single projects to large batches, but typically the variety is very high, while the flow of products is relatively low (HVLV).

Due to increasing competition from other companies, the company feels the need to improve its performance, mainly with respect to delivery reliability and lead times. Within the company it is believed that a lower lead time might attract more customers. Furthermore, the current way of planning is based largely on experience of employees and is not based on actual capacity data. Therefore, the organization lacks knowledge about which factors actually relate to performance. A few years ago, the company has started collecting more data from the shop floor, in order prepare for future solutions. To some extent the company has tried to implement some minor lean principles (e.g. order acceptance based on max. 85% utilization), however no major changes have taken place in the last years.

3.1.1Customers

The company does not serve many customers. In total, only 94 customers are served in one years’ time, with about an 80/20 distribution. Thus, 20 percent of the customers make up about 80% of all orders. This is an important observation, as this means that a large part of customers is frequently ordering. This creates opportunities for coordination of planning along with the customers, instead of situation in which customers are largely unique for every order (e.g. ship building industry). Therefore, the company is largely classified as a Repeat Business Customizer (RBC) and partly as Versatile Manufacturing Company (VMC). With RBC, customers require customized products over the length of a contract. With VMC, companies compete for individual orders (Amaro et al., 1999).

3.1.2 Shop floor configuration

(11)

3.2 PPC System

The production planning and control system of the company will be described following the five decisions points that are considered to be of crucial importance to the final delivery performance (Soepenberg et al., 2012a). These decision points are related to the control points of WLC. The decision/control points include customer enquiry, order acceptance, release, (priority) dispatching and output control.

3.2.1 Customer enquiry and acceptance

At the customer enquiry stage, usually the customer requests a specific delivery date. Customers of the company are usually OEM manufacturers that rely on a specific delivery date, as their inventory is often insufficient to store all parts needed. To check whether this delivery date is achievable, the planner checks if capacity is available. Capacity is measured in Euro’s over the whole process and via a quick Excel sheet it is determined if an order fits within the planning. The available capacity measured by a combination of available labour (hours) and average turnover per hour (euro’s). This capacity is based on the process as a whole and does not consider individual capacity groups. Delivery dates are thus not set by the company itself, but rather accepted or not. The company has adapted the lean principle of accepting to a maximum capacity of 85%. However, this fictive capacity is often violated as last-minute rush orders of important customers are accepted.

3.2.2 Pre-release

The company makes use of Master-production-schedule (MPS) that is generated by the ERP system. The MPS is the result of the bill-of-materials (BOM). Within this BOM, any necessary materials and operations are specified, as well as the prognosed set-up and operation times. The MPS uses a backward scheduling system (from the delivery date as requested by the customer), where the ERP-system determines the routing and timing of operations. A maximum of one operation per day is scheduled. Currently, the company calculates for 5 days as standard safety time to account for any uncertainties. This includes the (re-)design of the product and nesting opportunities, for example. A maximum of seven days is allowed in order to nest products as efficiently as possible. However, this causes a standard increase of lead time and also leads to excessive inventory on the shop floor. Nesting is discussed in detail in section 4.

3.2.3 Release

The release decision is of main focus in this study. In its current form, the release decision is controlled by the nesting of orders on raw materials. It is the responsibility of the “nester” to release nests to the cutting operation. The nester tries to fit orders within a time frame of seven days, however due to the high degree of variation of raw materials, this is not always possible. Generally, the nester will wait until he can make full advantage of a sheet of raw material, disrespecting the set seven-day timeframe regularly. (Material) cost-efficiency is an important aspect for the company and as such, orders are released to the shop-floor mainly by determined by nesting opportunities and not on release/due dates. In the past, the cutting operation was considered to be a bottleneck, and since then has been treated as such.

3.2.4 Dispatching

(12)

neglected. The central planning department only intervenes in case orders risk extreme lateness or when orders are of high priority come in.

3.2.5 Capacity

The press-brake operation is considered to be the bottleneck of the process. The average utilization of this operation is 93% and, in some periods, the utilization is over 100% by means of working overtime. In the current situation it is made sure that this operation does not suffer from starvation. This is partly the result from the current way of releasing jobs to the first operation, which generates great amount of inventory before the press-brake operation (often the second operation). Other work cells average about 70-75% utilization.

Since all but the first operations are DRC, the capacity also greatly depends on the available labour. The press braking operation, for example, has 14 available machines and only 12 assigned workers. Moreover, in the summer, a significant decrease in capacity is available as employees enjoy their holidays. However, DRC also has its upside, as the company is able to temporarily increase its capacity by working overtime, or deploying temporary workers, which is not uncommon practice.

3.3 Performance

Performance of the company is currently measured in quite detailed manner. de Man & Mannhardt (2019) introduced a performance measurement system within the company. It is argued that the company now has a relatively comprehensive performance measurement system for this type of company.

It is observed that delivery performance and lead time within the company are poor. Of all orders of 2019, it is observed that 72,85% is delivered early or on time (with 22% exactly on time) and 27,15% is delivered late. The average tardiness of orders that are late is 10 days with a standard deviation of 14,95 days. The lateness is calculated by subtracting the packing date from the promised delivery date. See Figure 7 for a histogram of the lateness distribution. The average lead time is 26,3 days with a standard deviation of 23,5 days. This considered to be very high, as the average process throughput time per operation is one day. An order with 6 operations had therefore a throughput time of 6 days. The distribution of lead times is visualized in Figure 5. Measuring lead time is relevant in this case, as company goals are to shorten the time between the placement and delivery of an order. The company goal is to improve delivery performance to 95% and reduce the average lead time to a maximum of 8 days by 2021.

(13)

Figure 4: utilization per capacity group for 2019

Figure 5: Lead time distribution

3.4 Root cause analysis: release/nesting

Delivery performance is becoming increasingly important for SME’s (Soepenberg et al., 2012a). Thürer, Stevenson, Land, et al. (2019) state that delivery performance is even the most important performance indicator for an MTO job shop. It has already become evident that the delivery performance of the company is relatively poor. The delivery performance is further analysed following the framework of Soepenberg et al. (2012a). The purpose of this analysis is to determine the root cause(s) underlying the poor delivery performance. The framework distinguishes between average and variance oriented diagnosis which relate to two performance objectives that together determine delivery performance: throughput improvement (reduces average lateness) and timing (reduces variance of lateness)(Martin Land, 2006). The result of this analysis will indicate to which of the five decision points (customer enquiry, order acceptance, release, dispatching or output control) the cause relates. The identified cause(s) will further determine the focus of the study.

0,0%

20,0%

40,0%

60,0%

80,0%

100,0%

120,0%

Jan

Feb Mar Apr May Jun

Jul

Aug Sep

Oct Nov Dec

u

tiliz

ation

Month

Utilization Capacity Groups

lasercutting/punching

pressbrake

pressing/machining

Welding/finishing

0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 120,00% 0 200 400 600 800 1000 1200 1400 1600 0 5 ₁₀ ₁₅ ₂₀ ₂₅ ₃₀ ₃₅ ₄₀ ₄₅ ₅₀ ₅₅ ₆₀ ₆₅ ₇₀ ₇₅ ₈₀ ₈₅ ₉₀ ₉₅ 10 0 Mo re Fr e q u e n cy Days

Histogram lead time

(14)

The first step is to analyse distribution of lateness. Similar to Figure 6b, it is observed that the lateness (Figure 7) over all orders is of high variance. Therefore, the analysis starts variance oriented.

Figure 6: (Soepenberg et al., 2012a)

Figure 7: Lateness distribution

Differences among subsets are analysed. The observed variance cannot be related to specific order sub-sets. Therefore, variance-oriented diagnosis is continued. The next step involves analysing differences over time. It is observed that lateness cannot be attributed to specific periods in time. It must be noted that at the beginning of the year (January) the delivery performance seems best, with relative few late delivered orders. However, this is explained by the fact that in the provided data, no sales data from 2018 is present. Therefore, lateness can only be calculated for orders placed after 1-1-2019. For the remainder of the year, variance is observed in every month. Therefore, it is assumed that the variance is not related to specific periods. Now it must be determined whether to continue in delivery time promising process (DP) or realization process (RP). This is done by means of order progress diagrams. Soepenberg et al. (2012a) distinguishes between two situations: (1) the variation of delivery dates is low at the start of the process and increases during the manufacturing process, asking for a focus on the RP (Figure 8c) or (2) the variation is very high to begin with and is straightened out by the manufacturing process, asking for a DP focus (Figure 8d). From the data available however, it is not possible to construct such progress diagrams. Whilst quite detailed production data is available, data that serves the tracing of individual production orders throughout the process is absent and therefore no distinction can be made between different batches of the same order. However, during a company visit it became clear that the variation in delivery dates was highest in the first operation and that every subsequent operation tries to improve this performance by selecting an order with the earliest delivery date, straightening out the variation as in Figure 8d. This was also presented visually by means of internal delivery performance diagrams. Unfortunately, this data could not be exported out of the ERP-system. However, it is sufficiently evident that the company generally improves on delivery performance during the process.

0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 120,00% 0 500 1000 1500 2000 -25 -23 -21 -19 -17 -15 -13 -11 -9 -7 -5 -3 -1 1 3 5 7 9 11 13 15 17 19 21 23 25 Fr e q u e n cy

Days early - late

Histogram lateness

(15)

If the framework of Soepenberg et al. (2012a) were to be followed, this “straightening of delivery performance” would suggest a focus at the DP process, as the RP is able to function properly. However, this direction might not be desirable as influence on delivery dates is limited. Moreover, the process cannot be classified either as reducing variance (Figure 8c) or increasing variance (Figure 8d) throughout the process. Rather, the process shows to be a combination of both. Estimated lateness is low at the beginning of the process, because of added safety time. At the release, great variance is introduced to the system because of the combined release of orders with different due dates via nesting. Finally, the manufacturing process tries to straighten out the variation by selecting orders with the earliest due date first. This also answers the next question from the analysis: is variance increased at the release of orders to the shop-floor? The answer to this question is that this is indeed true. Therefore, the variance-oriented analysis is continued.

Figure 8: Progress diagram examples (Soepenberg et al., 2012a)

The final question that will lead to the root cause is: is variance increased between the completion of the last pre-shop floor process and release? To this the answer is both yes and no. The release of the nests is both determined by a planning schedule and availability of machines. The current release method is controlled by nesting and it has shown from practice that the process of nesting introduces a mix of orders with various delivery dates to the shop floor. Normally, a nest is created a few weeks in advance and a planned release date is set. However, it also happens that opportunities arise to release a nest early, because machines are idle. Then, the nester releases the nests earlier to the system than planned. Also, it might take longer to fill a full sheet of raw material. In that case, the nesters usually wait until he can optimally make use of the raw material, resulting in a delay of all orders within that nest. Thus, either the release follows it planned release date or it is released manually by the nester.

Therefore, it is concluded that the process of nesting orders is the main cause of introducing variance to the system. Since the cost-effectiveness that lies behind the idea of nesting cannot be neglected, nesting must continue to take place in some form. If WLC would be implemented in its regular form, the system would not have (complete) freedom to determine optimal release decisions. It is argued that a solution must be designed that connects nesting with the release decision, apart from any other complexities that are to be identified and constrain WLC functions.

(16)

3.5 Complexities and Issues

This paper distinguishes between complexities and issues. Complexities are the characteristics of the company and the market it operates in and are regarded as a “given” element that cannot be changed. Thus, complexities cannot be solved, rather, they need to be incorporated in a future PPC system design. Issues, on the other hand, are sometimes resulting of these complexities, but can also exist independently (without connection to a complexity). Issues are observed from practice (e.g. poor delivery performance), however, they have the potential to be improved upon. The relationship between complexities and issues are visualized in Figure 9. Thus, during the PPC system design, extra care should go out to take into account the complexities and solving issues at hand.

Figure 9: relationship between complexities and issues

3.5.1 Complexities

The focus of this research is to address the complexities (1) nesting and (2) DRC. Additionally, three other complexities are observed that complicate the applicability of WLC. These complexities have less impact as nesting and DRC; however, they cannot be neglected in the final systems design.

1. Nesting. From the root-cause analysis as from Soepenberg et al. (2012a) it has showed that nesting is the main source of delivery date variance in the process. As nesting is important for the cost-efficiency of raw materials it cannot be discarded. Nesting has to be considered by the PPC system to enjoy the cost-efficiency benefits. Currently, the nesting process determines which orders will actually be released. One employee (the “nester”) performs the nesting of orders by manually designing nests using specific software. Once orders are accepted and have been translated into technical specifications, the order becomes available for the nester. He will, based on experience, create nests of orders for the few oncoming week (up to 4 weeks in advance). The aim is to create nests of orders with delivery dates with a range of 5 days. Once a nest had been made, it is not reconsidered again. However, due to the variety of material requirements, it might happen that orders with a great spread of delivery dates are combined into a nest, resulting in a great variance of delivery dates that are released to the shop floor. 2. DRC. The capacity is not only constrained by machine capacity, but also by labour capacity.

“There is very limited research on the performance impact of order release control methods in

DRC shops. Moreover, this small body of work does not consider recent advances in the order release literature and typically focuses on simple forward/backward loading based release mechanisms only” (Thürer, Stevenson, & Renna, 2019). DRC is especially prevalent at the

bottleneck operation, as this process requires continuous labour in order to operate. If labour capacity can be considered by a PPC system, this would ensure the optimal deployment of employees.

Delivery performance

(17)

3. Unequal specifications of machines. Whilst it may seem that the process of the case company is relatively simple with just 6 operations, each operation may include up to 14 different machines that fulfil the same function. However, the machines are unequal in their specifications, especially at the bottleneck operation. Some machines are smaller than others, or not suitable for some specific pressings. Also, more modern machines are present with a greater throughput rate than others. This makes it difficult to assign a job to each machine, as its impact on future workload is unknown. For example, an operator might choose to use a specific press-brake only because it has a high throughput rate, whilst the upcoming order is specifically assigned to that same press-brake as it is over 4,00 meters long.

4. Sequential batching. Different orders require different tooling. This is mostly of concern to the bottleneck. The current system does not consider sequential batching as there is no information on required tooling. Orders at the shop-floor are selected by worker experience in order to minimize set-up times. The future PPC system should therefore consider the sequence dependency between orders to avoid unnecessary set-up times at the bottleneck operation, if possible.

5. Limited influence on delivery dates. A large share of customers is frequently returning (RBC) and have become used to request delivery dates themselves. Because of space limitation and other synchronisation issues, it is often not possible to deliver on other days. This is greatly limiting the ability of the company to propose delivery dates themselves, and thus so in realising these delivery dates. Therefore, there is more dependency on order acceptance, release and dispatching in order to fulfil the requested delivery date.

3.5.2 Issues

Aside from the complexities that are identified in the previous section, several issues are observed within the case company. These issues are somewhat more case-specific; however, they are related to some of the complexities and mechanisms of the system. The identified issues are given in Table 1.

Issue Explanation Related

complexity

Also confirmed by Long and variable

waiting times after cutting (big buffer)

Caused by the order release not considering workload at the capacity groups. WIP is not limited.

Nesting (de Man & Mannhardt, 2019)

(Wienholts, 2020)

Poor delivery performance

Is partly a result from the different machining centers focus on optimizing the utilization of their bottleneck machines. Reasons for this is the already high utilization due to labour constraints (DRC). Also, the set-up times are minimalized by selecting order that do not require set-up. These factors are not considered at

release/nesting. Nesting DRC Sequential batching Unequal machines

(de Man & Mannhardt, 2019) (Wienholts, 2020)

Operator freedom to select orders themselves

“Easy” orders and orders from important customers have higher chance of being selected, enjoying a much higher delivery performance.

Set-up times (de Man & Mannhardt, 2019) (Wienholts, 2020)

The planning method based on revenue capacity does not account for bottleneck operations.

At order acceptance, capacity is calculated in euro’s as a whole for the total process. Possible workload to different capacity groups is not considered at acceptance. No real effort on the realistic estimation of delivery dates is undertaken.

As a result, the bottleneck is frequently overloaded.

Limited influence on delivery dates

(de Man & Mannhardt, 2019) (Wienholts, 2020)

(18)

During the research process, it was observed that the quality of data was relatively poor, despite efforts to improve information quality over the past years. Data is collected by operators that enter their respective working hours at each machine. The first operation (cutting) is able to work largely unmanned. Operators are only required for (un)loading and setting up the machines. Therefore, no reliable operating times are collected for this operation. This is evident from the data, as a large part of orders are processed in unrealistic amounts of time. “Incomplete and thereby compromised data, if used

for future scheduling, will lead to an unrealistic or even false production schedule and in the end to a possible violation of promised delivery dates. Hence, measures to prevent collecting corrupted data or resolve errors after the gathering are necessary to increase the accuracy of the decision making in PPC.”(Schuh, Reuter, Prote, Brambring, & Ays, 2017). Therefore, the quality of data should be

(19)

4. Theoretical background and discussion

It has been widely acknowledged that complexities of real-life job shops present difficulties when developing a robust solution such as WLC (e.g. Cransberg et al., 2016; Mark Stevenson, Huang, Hendry, & Soepenberg, 2011). Therefore, this section discusses the identified complexities in relation to the functions of WLC.

4.1 Workload Control: release, dispatching and capacity adjustments

The root cause of poor delivery performance was traced back to the release decision in section 3.4. To indicate the relevancy of combining nesting into WLC, especially at the release decision, it is important to know that: “only the release decision can ultimately prevent workloads from building up on the shop

floor and only dispatching decisions can correct for the inevitable uncertainty incurred during the progress of orders on the shop floor” (Cransberg et al., 2016).

The release decision of WLC performs two basic functions: load balancing and timing. These functions are fulfilled by two phases within the release procedure: selecting and sequencing, respectively. The timing function tries to satisfy the planned release dates by sequencing the orders by the earliest planned release date first. The load balancing function ensures that workload norms of capacity groups are not exceeded, by only selecting orders that do not violate these norms (Mj Land, 2004). To become familiar with the release function, a popular method of releasing is explained: LUMSCOR (Lancaster University Management School Corrected Order Release). The principle is as follows: before the actual process of release takes place, orders are scheduled backwards from their delivery date. Orders that are in the pre-shop pool are sorted according to their planned release date. The order with the earliest planned release date is considered for release first. For the routing of this order, its processing time at each operation is compared to the workload of the corresponding station. If the norms of workload at each station are not exceeded, the order is staged for release and its workload is added to that of each operation. If norms are exceeded, the next order is considered and the current job is left in the pshop pool and re-considered at the next release moment. What is special to LUMSCOR, is that it only considers the workload for a given operation the time that an order spends there. (Thürer, Stevenson, & Renna, 2019). In traditional WLC, it is argued that simple dispatching rules should be sufficient to adhere to the set due dates if the release decision is performed properly (Henrich et al., 2004). However, if no optimal release decision can made due to complexities (e.g. nesting) the importance of other control mechanisms (such as capacity adjustments) increases in order to deliver good performance (Cransberg et al., 2016; Thürer, Stevenson, Land, et al., 2019). Moreover, capacity adjustments have shown to have a positive influence on lead time (Thürer, Stevenson, Land, et al., 2019). Therefore, it is argued that dispatching decisions and capacity adjustments should receive extra attention in the design of a solution.

4.2 Complexities

One of the pitfalls of previous WLC literature is that the concept is often developed with little connection with reality (Mark Stevenson et al., 2011), hence the large amount of simulation studies that contributed to the refinement of the concept. It is not until recently, that literature has shifted its focus towards the identification of real-life complexities, dynamics and aspects that make the WLC concept difficult to implement successfully in practice (e.g. Cransberg et al., 2016; Soepenberg et al., 2012c, 2012b; Mark Stevenson et al., 2011).

4.2.1 Nesting

To understand how nesting can be incorporated into WLC, one must first understand the basic principles of nesting.

Definition: “Nesting combines multiple orders that require the same type of sheet metal so that a punch

(20)

as it “uses larger, standard sheets, which cost less per pound” (Herrmann & Delalio, 2001). The main goal of nesting itself, is to address costs for direct labour that is associated with setting up machines and the cost of materials. Typically, the nesting in the sheet metal processing industry is based on two important aspects: material type and thickness. To form a nest, the geometry of a part and the quantity that has to be produced is of importance to determine the allocation (Herrmann & Delalio, 2001). In their paper, Herrmann & Delalio (2001) present an algorithm that dynamically determines “good” nests. This heuristic calculates the area needed for order on a sheet of raw material and can run independently from any other software packages. According to the case company, current nesting software does not support such dynamic heuristic; it only allows for optimal and exact allocation of parts onto a sheet of raw material. This process takes time, and therefore nests are not reconsidered (i.e. dynamically) once they have been made.

Compared to the WLC release method, one can observe that the aspects of nesting and release decision differ very much. Where a good load-balancing WLC system will only need a relative small pool of orders to select from (Mj Land, 2004), the pool size needed to form nests will depend on the variation in material type and thickness requirements of orders (Herrmann & Delalio, 2001). As variety is high in a HVLV market, the chance of needing a large pool are real. For example, the case company currently considers orders with a DD spread of 7 days for nesting.

In the current situation, the release of a nest puts a mix of orders with various delivery dates on the shop floor and thus, for most of these orders, their planned release date was not realized. This is important, since disturbing the planning release sequence negatively influences the timing performance (Mj Land, 2004) and thus delivery performance. Moreover, in practice, nests are created many days or even weeks in advance of their planned release date and are then “frozen” until actual release to the system. WLC on the other hand, determines the exact selection of orders at the moment of release.

As most WLC systems are rule-based, they rely on some fixed parameters (e.g. workload norms). This makes it inherently difficult for WLC systems to adapt to the dynamic and complex environment in which modern companies operate (Missbauer & Uzsoy, 2020). From the perspective of nesting, it will be difficult to determine whether a more optimal nest should be released, possible exceeding workload norms, or to go with a sub-optimal nest, adhering to the norms. Also, large nests with a lot of the same parts will have less chance of being selected, as they exceed workload norms for specific capacity groups. This problem is somewhat related to the timing problems that are mentioned by Mj Land (2004): non-urgent order might be selected as they fit workload norms, or very large orders may be postponed as they never fit workload norms. This suggests that with a large order pool (as needed for nesting), there is greater chance of selecting non-urgent orders and so, the planned release date is not adhered to more often. As a solution, the periods between release should be long enough, so that more urgent orders with higher workloads have a chance to become selected. A second solution is to define looser criterion for larger orders (Mj Land, 2004), or nests in this case.

Therefore, it is argued that with current nesting practices, there is no possibility to consider load balancing or timing at the moment a nest is created. And so, it is argued that nesting and release considerations should take place at the same time, in order to combine their aspects.

An example of a PPC system that does considers nesting, is the study of Slomp, Bokhorst, & Germs (2009). Here, the planner is made responsible to fill a shop buffer in order to balance loads over time. However, in this example, nesting is still performed manually and thus is not considered dynamically at the moment of release.

(21)

the first operation step, whilst DRC relates to the bottleneck operation, which is the second operation. Also, the criticality of both complexities is considered to be very high. The framework points towards the release decision as appropriate control measure and thus confirms that a focus on release when considering DRC and nesting complexities is appropriate.

Figure 10: Position and criticality of complexities (Cransberg et al., 2016)

Moreover, “it becomes necessary to look beyond managing the two-dimensional (2D) trade- off between

the timing and workload balancing functions” (Cransberg et al., 2016). This indeed seems to be true, as

the cost-efficiency of nesting becomes directly connected with the release function of WLC, suggesting a third dimension on the trade-off. Cransberg et al. (2016) suggest further exploration of companies where nesting is part of the process. This makes this study particularly relevant. However, some suggestions are still relevant. The suggestion is to consider sub-optimal release decisions if dispatching allows for sequencing improvements. Also, buffers that are added to the system to allow for the re-sequencing of orders, can also be used to address other complexities (Cransberg et al., 2016).

4.2.2 Dual Resource Constrained

Stevenson et al. (2011) pointed out that many WLC simulation studies do not consider capacity to be constrained by both machines and labour (DRC) and call for more research on WLC in DRC situations. Even now, literature on order release mechanisms in combination with DRC is scarce (Thürer, Stevenson, & Renna, 2019), which shows from the relative small size of literature research on DRC by Thürer (2018). Research of Thürer, Stevenson, & Renna (2019) shows that WLC has great potential in DRC shops. However, they assume complete interchangeability of workers, and as is observed at the case company, this is not the case in reality. Moreover, the focus of their research was to incorporate information with respect to labour availability such as absenteeism, which is a different focus than this research.

As has been described in section 3, the company is limited by the number of operators that it employs. At the bottleneck operation, there are more machines than operators. Therefore, the release decision should consider labour capacity, as it cannot release orders that have workload on all machines simultaneously. Arguable, these DRC restrictions can be implemented relatively easy by just defining capacity norms in terms of labour capacity.

4.3 Summarizing

(22)

1. Nesting and the release decision should be made at the same moment, so that timing and workload balancing functions can be performed at the same time as optimal material usage. In order to maximize opportunities for nesting, the nesting of orders should be performed at the moment of release. Some mechanism must be in place in order to determine the balance between load balancing, timing and material efficiency.

2. The periods between release should be specified to allow for different size nests to be considered for release. A mechanism must be in place that prevents nests being neglected that do not provide the most even load balancing (particularly relevant for large orders). This also concerns the consideration between sub-optimal nests and workload norms. It might be beneficial to release a nest that is very optimal and exceed norms slightly. Capacity adjustments can then possibly address the extra workload.

3. DRC norms. At the moment of release, DRC aspects should be considered also. Whilst an optimal nest might be created that also does not exceed machine workload norms, it might conflict with available labour. DRC should therefore be incorporated into the norms that prevent undesirable releases.

(23)

5. Proposed system: Dynamic Workload Control

In the previous section, literature supported the forming of mechanisms that integrate nesting and DRC into a WLC oriented system. The focus of the design is to address mainly nesting and DRC complexities, however, the proposed system covers the complete process from order entry to order exit, in order to not waste any lessons learned during the research process. In the designed system, several elements of traditional WLC are retained and therefore explained in less detail. In Table 2 the conflicting complexities and issues are related to WLC levels and PPC decision points. For each decision point, a suggestion is given on how the future system should handle the complexities and issues. The suggestions are then explained in detail. A visual representation of the system can be found in Figure 11.

Figure 11: proposed system

5.1 Order entry

As due date setting (by the company itself) has a major impact on the variance of lateness (Thürer, Stevenson, Land, et al., 2019), it is suggested that the company examines the possibility to set the final

Table 2: systems design suggestions

WLC level Decision points

Should address the complexity

And solve the issue of

By following the suggestion: Order

entry level

DD promising Limited influence on delivery dates

Unrealistic Delivery Dates

Minimize customer influence on DD’s and suggest DD based on information from the system.

Realistically set DD’s are easier to adhere to. Order

acceptance

Maximize order acceptance of orders that may complete nests using a dashboard that indicates incomplete nests.

Minimize order acceptance of orders that conflict with workload norms

Order release level Release Nesting DRC Inequality of machines Sequential batching Poor delivery performance Poor load balancing Long lead times Excess WIP

Simultaneous Nesting and Release using an algorithm

Decision making in sub-optimal situations

Order dispatching level (priority) Dispatching Sequence dependant set-ups Opportunity behaviour by workers

Minor buffers between operations allow for the re-sequencing of orders in order to achieve benefits on combining orders with sequence-dependant setups.

All levels Output control DRC High utilization of bottleneck

Consider capacity increase to enable release of large orders.

(24)

delivery date themselves. The larger part of orders will pass this stage and directly flow to the order acceptance stage, as delivery dates are mostly enforced by customers. If it is possible to set a delivery date by the company itself, there is a greater chance that that delivery date will be realistic. The system should provide the sales department with an overview of planned workloads for each capacity group so that they can estimate the impact on the system if the order would to be accepted. Also, information on nesting should be made insightful to the sales department, in a way that they can accept orders with specific material requirements in order to complete nests. These functions become especially important when the ability to influence delivery dates increases, which might differ per company or market. At the acceptance stage, the main criteria for accepting or rejecting an order should be the long-term utilisation of the process. Assuming that all accepted orders are to be manufactured, order acceptance or rejection is the only way to control long-term utilisation levels (Mj Land, 2004). Of second interest is the expected workload on the bottleneck. The system provides information on the workload of planned releases and therefore, the sales department should not take in orders that require workload on already overloaded capacity groups. Overloading the bottleneck would severely impact the throughput time of an order.

5.2 Assigning attributes

Directly after the acceptance of an order, attributes should be assigned to the order. The purpose of these attributes is to facilitate nesting and release decisions by the system. These attributes include of course the planned release date (determined by backwards scheduling of operations from the delivery date) but also the attributes for nesting, such as material, thickness, surface area, quantity and optionally the geometry. Also, additional attributes that relate to the workload of the bottleneck operation should be added, such as an indication at which machine the parts of that order can be produced. The latter can be used to not only provide load balancing on machines (because of inequal specifications of machines) but also to determine labour requirements (the number of machines that need to be operated). In order to not make the system overly complex, attributes of other operations than the bottleneck are not considered at this stage. The attributes are summarized in Table 3. As soon as attributes are assigned to each order, the order moves to a pool of orders.

Concerns Attribute Function

Nesting Material type Determine raw material requirements

Thickness

Surface area Estimate total amount of raw material Quantity

Geometry (optional) Exact forming of a nest

WLC - timing Final delivery date Determine planned release date by backward scheduling (1 operation = 1 day)

Number of operations WLC - load balancing DRC Estimated processing time

Determine workload per individual machine at the bottleneck. Also, the required amount of labour capacity can be calculated

Machine requirements

Table 3: Attributes

5.3 Pool of orders

(25)

select orders from that provide a balanced workload to the system. However, in this situation, the pool also serves the nesting process by providing a variety of orders that can be selected to form a nest. As soon as the attributes (Table 3) are added to the order, the system can start to calculate possible nesting and releasing options. From these attributes, a relative accurate estimation can be made on the workload it will add to specific machines at the bottleneck when the order is to be released. At the same time, nesting can be considered. For this, an algorithm is needed that allocates orders to raw materials, based on the material type, thickness area and quantity. Ideally, this algorithm directly serves the task of actually forming the nests and its output is directly interpretable by the cutting operation. For this the exact geometry (technical drawings) should be included. However, such algorithm is considered to be very complex, which is not in line with the aim to keep the system relatively simple and robust. Simpler algorithms exist that provide an estimation on nesting possibilities based on the surface area e.g. (Herrmann & Delalio, 2001). The actual nesting should than be performed by a worker and, since the system has already considered the nest to be the most optimal release at that moment, this process can take place just before actual release (and not days/weeks earlier as in the current situation).

5.4 Release

Next, the first release decision is made. This is the so-called “rough-release” as it does not consider the complete routing of every order and workload for every capacity group in the process. The only purpose for this release is to balance load on the bottleneck operation. This is somewhat different to traditional WLC, where the main objective of the release is to provide load balancing to all capacity groups. The reason for this being that the focus of the system is to make optimal use of the bottleneck, as that will determine the total throughput of the system. Between the time of initial release of an order, after which it is processed at the first operation and then spends time in the shop-floor buffer, a lot of uncertainties may have happened; e.g. workers have become unavailable (DRC), machines might have broken down, delays may have formed or rush orders might have been forced into production. Therefore, it does not seem logical to account for all this variation at the release, several days before the orders actually reach the bottleneck. Moreover, at subsequent operations there is less average utilization and capacity can be increased more easily, therefore being less critical.

The optimal selection of orders that are to be nested together (and thus released together), is continuously re-calculated by the system using the attributes that were assigned to each order. This way, the system can optimally make use of all orders that are within the pre-shop pool. An algorithm should determine the best trade-off between load balancing, timing and cost-efficiency. The exact specification of this algorithm will differ per company, as each company will value these aspects differently. By maximizing the time between order entry and nesting (by performing nesting at the latest moment), there is a broader selection of orders from which nests can be formed, positively influencing the cost-efficiency of nests. This will become even more evident when total throughput times start to decrease (e.g. by discarding the 5 days of added safety time), because in that case, less time (and so less orders) is available to select nests from. Also, opportunities arise to include rush orders that come in last-minute, which normally could not be added to already defined nests.

In addition, the release considers norms that are set for labour capacity using the attributes. For example, a norm may be set that a maximum of 12 machines can operate simultaneously at the bottleneck, ensuring that no shortage of labour will occur. The exact specification of norms will depend on specific cases and therefore is not included in the design of the system.

5.5 Shop floor buffer

(26)

status of this shop-floor buffer in order to prevent excess inventory. This can be incorporated into the system by setting a norm on the maximum size of this buffer. From this buffer, orders will be reselected and re-sequenced in order to be released.

5.6 Re-release

In traditional WLC concepts, all operations are controlled only by dispatching and capacity adjustments, once orders are released to the shop floor. This means great dependency on quality of the initial release decision to the shop floor. Moreover, “considering a complexity may lead to undesirable consequences

at other work centres both upstream and downstream of the complexity if the ideal sequences at those resources differ from that where the complexity occurs. .. This makes it less obvious that all complexities should be considered simultaneously at release” (Cransberg et al., 2016). As a safety mechanism, a

second releasing function is proposed, that reconsiders the sequence of orders and determines exact machine allocation for the remainder of the process. This re-release will function exactly as in traditional WLC concepts (e.g. LUMSCOR). It will sequence orders based on their planned release data and select orders that do not violate workload norms on the remaining operations.

The release decision to the bottleneck is made with the most up-to-date information available, taking full advantage of any changes that might have happened since the initial release. If a machine would break down, operators became absent or rush orders were released, the system can now recalculate the optimal sequence of orders to the bottleneck. Another complexity that the re-release addresses is that now the sequence dependency related to set-up times will be considered, in line with the suggestion to not consider all complexities at once. As a side effect, re-release addresses the issue of opportunity selection of orders by operators, as the system will now determine which orders in which sequence are to be processed. Hereby ensuring that load balancing is not put into danger by poor operator decisions. Moreover, utilization is improved as set-up times are minimized.

In order to prevent orders that were initially released very early keep waiting in the buffer (due to nesting), a maximum waiting time should be determined. This might cause the order to be processed early at all subsequent operations and result in a greater inventory of finished goods. However, it prevents the buffer from exploding in size and ensures a short throughput time as an optimal workload balance. Also, it minimizes the risk that the order is slowed down further downstream the process.

5.7 Dispatching

Once the orders are processed by the bottleneck operation, simpler dispatching rules take over. To reduce the variance of lateness, due date oriented dispatching rules provide the best performance (Mj Land, 2004) such as earliest due date (EDD) or modified operation due date (MODD). However, the exact selection of a dispatching rule is out of the scope of this study. It is however important to enforce adherence to this dispatching rule among, as previously has been observed that workers are susceptible to select orders that are easier to process first. Although this is a case-specific recommendation, it can of course apply in every company were operators are given freedom to select orders themselves.

5.8 Output control

(27)

(28)

6. Validation

This section is dedicated to convince the reader of the pragmatic validity of this study. At the basis of pragmatic validity in design science research is the field testing of the proposed design (van Aken et al., 2016). Due to time constraints that apply to this thesis, an actual implementation of the system was unfortunately not possible. Moreover, it must be noted that this research was conducted will the world was suffering from the Corona-crises, limiting access to the case company. However, discussing the design with field experts and practitioners is possibly the next best option as it can provide pragmatic validity and practical (van Aken et al., 2016). Also, an extensive body of literature was used in order to support the design of the system.

The proposed design of this study is validated at the case company by means of an in-depth discussion with several of the company’s employees. These include the general manager, operations manager, planner and three internship students that studied the system. All of them have in-depth understanding of the system at the company and moreover, have experience at other, similar companies. In general, they very much liked to proposed solution. The general manager has worked for several other metal sheet manufacturers and can be considered an expert in the field. The manager confirmed that the complexities indicated in this study (e.g. limited influence on due date, nesting, and DRC) are very common amongst other metal sheet processing firms, confirming that the design deals with an authentic field problem. That complexities (included nesting) in job-shops are authentic field problems was already confirmed by (Cransberg et al., 2016).

The employees agreed that the system that continuously calculates the possible nesting of orders is very much of interest. To their knowledge, there is no existence of a system that integrates the nesting of orders into a planning system. The idea that nesting aspects need to be considered at the same time as the balancing of workload, has already come to mind of the manager. However, it never came to an exact specification on the mechanisms of such system. This study shows that by assigning attributes in combination with the continuous calculation of possible nesting options, such system starts to take form. Also, the second releasing function, that reconsiders the sequence and selection of orders from the shop-floor buffer, serves as additional safety mechanism that copes with any uncertainties that characterize the HV/LV industry. The manager has indicated that further development and implementation of the system within the company is much desired.

Improving Workload Control applicability: mechanisms to address real-life complexities