I
Working with Loads
Introducing Workload Control at SPX Process Equipment BV NL
P.L. Bruinsma
June 2010
II
Working with Loads
Introducing Workload Control at SPX Process Equipment BV NL
Master´s thesis
Technology Management
Company:
SPX Process Equipment NL BV
University:
Rijksuniversiteit Groningen
Study:
Technology Management
First supervisor:
prof. dr. ir. J. Slomp
Student:
Peter Laurens Bruinsma
I
PREFACE
This thesis is the final obstacle of the course program of the Master Technology Management at the University of Groningen. During this course program my interest in internal business logistics increased and after having attended a workshop in lean manufacturing, provided by the Lean Operations Research Center (LO-RC), I decided to perform research in a logistics-related area. The LO-RC offered me a research project at SPX Process Equipment NL BV. My internship at this company was a great experience and allowed me to execute an independent research assignment. In this preface I would like to thank those who enabled me to achieve this milestone.
I would like to start with special thanks to the supervisors who supported me during my research. My first supervisor, prof. dr. ir. Jannes Slomp, gave me, as a member of the LO-RC, the opportunity to get access to high-level scientific literature in the field of lean manufacturing. Besides, his knowledge and feedback provided me new insights when I lost focus and our discussions largely contributed to this final result. Also I would like to thank my second supervisor, dr. Nicky van Foreest, for helping me structuring my research and providing useful comments for the final version. I wish to thank my company supervisor, Robert Weerkamp, as well. His time and patience made me feel comfortable within the company from the very beginning and the resources he provided me, accelerated the progress of my research.
At last I want to thank my family and friends for their help during the research. Especially my parents, who dragged me through the whole trajectory with their emotional and financial support, and my girlfriend Marinte, who gave me positive energy at moments of despair.
My aim was to produce a thesis that is readable for a broad audience, and in that sense I am proud of the final result.
Zwolle, June 2010
II
ABSTRACT
This thesis presents a case study at SPX Process Equipment NL BV, a company that produces a wide variety of centrifugal pumps within the range of standard to highly customer specific products. SPX NL pursues a make-to-order (MTO) strategy in a production environment that can be characterized as high variety, low volume. Despite the lean manufacturing program that was introduced in 2006, the company faces major problems in meeting just-in-time (JIT) deliveries. With an average delivery performance of only 64 percent, the central research question of this thesis is:
How can SPX NL control her material in such a way that delivery performance will increase to 85 percent?
The thread of this research is an analysis of the influence of the production process characteristics and the current production planning and control (PPC) system on the delivery performance of SPX NL. A value stream analysis was executed to identify sources of waste in the material and information flow. The most striking source of waste was evaluated from a PPC perspective and unraveled in order to become the foundation of a redesign of the PPC system. The proposed PPC system is expected to increase the company’s delivery performance to such a level that the research question is answered.
SPX NL pursues a make-to-order (MTO) strategy in a high variety, low volume production environment. The manufacturing process is characterized as a general flow shop, in which a functional shop floor configuration facilitates large product variety. Since MTO companies focus on realizing short throughput times in order to gain competitive advantage, their PPC system should focus on the authorization of starting jobs. The current PPC system of SPX NL can be characterized as a push-oriented system with material requirements planning (MRP) used to release jobs. MRP systems behave like open queuing network in which jobs are released without regarding the number of jobs in the system.
The value stream analysis that was executed during the diagnosis phase shows large and fluctuating order congestions in front of the testing department, a work station that was already defined as the bottleneck station during previous research. The implementation of a planning board at the planning department revealed how the current MRP system affects the order throughput at the test bed. The two outcomes were:
The sales department accepts orders and issues delivery dates to customers according to expected lead times, without taking capacity of the shop floor into account.
The planning department releases orders according to the backward infinite loading (BIL) technique without taking capacity of the shop floor into account.
III
The MRP system does not contain a control element which facilitates coordination between the sales department, the planning department and the production department, whereby the push-based character causes major problems for test operators in meeting ODDs. The uncontrolled and fluctuating pattern of order arrivals results in on average 40 percent of orders that do not meet the ODD at the test bed.
The proposed redesign of the PPC system, consisting of a workload control (WLC) system and an order progress control system, is a load oriented control policy intended to establish short and precisely predictable lead times. It is aimed at keeping the workload in front of the test bed constant. Dividing the test bed into three capacity groups, each representing a group machines with the same testing characteristics, is considered as a condition for balancing the order stream towards the bottleneck station.
The WLC system consists of input and output control decisions at three hierarchical levels, all related to different phases in the order flow. The input control decisions are founded on the capacity restrictions of the test bed, expressed in the amount of available man hours. The three hierarchical levels are:
Order entry. This decision function is concerned with whether an order can be accepted and which delivery date should be issued. An order acceptation procedure is designed to support the sales department in accepting orders and setting delivery dates according to the capacity of the test bed. This procedure is founded on a capacity plan that consists of the weekly amount of available man hours per capacity group.
Order release. This decision function should control the amount of workload on the shop floor and attain a constant amount of WIP in front of the test bed. Setting workload norms for each capacity group should guide the planning department in following the order release procedure. The order release decision is a daily repetition of choosing such an order mix that the load level of each capacity group will not be exceeded and the latest release date (LRD) of orders will not be passed.
Order dispatch. This has become a modest decision function, since the load balancing function of the WLC system reduces queue lengths and hence limits the choice among jobs within the test bed. Therefore the earliest due date (EDD) stays the priority rule in order dispatching.
The output control decision functions stem from feedback signals from the different hierarchical levels in the WLC system. These signals can be can be divided in two loops:
Short-term feedback loops. Locating the feedback point (FBP) directly after a capacity group of the test bed provides the planning department with accurate and timely feedback information about the progress of individual orders. This, in turn, enhances the controllability of the shop floor, since the authorization signal for new order releases is founded on this information. It can also be used to determine when and where short-term capacity adjustments are needed. Possibilities for short-term capacity adjustments are cancellation of low-priority orders and overtime working.
IV
orders, rework percentage, amount of overtime hours and other delays. The output control decision is concerned with the determination of medium-term capacity adjustments for the next period. Changing the order acceptation mix and transferring internal test setups to other workstations are seen as the most effective way adjust the capacity of the test bed in medium term.
1
TABLE OF CONTENTS
TABLE OF CONTENTS ... 1
1. INTRODUCTION ... 4
1.1 SPX Process Equipment ... 4 1.2 Products ... 4 1.3 Manufacturing process ... 6 1.3.1 Mechanical department ... 7 1.3.2 Assembling ... 7 1.3.3 Testing... 7 1.3.4 Painting ... 92. RESEARCH DESIGN ...10
2.1 Initial problem statement and research question ... 10
2.2 Conceptual model... 11 2.3 Sub-questions ... 11 2.4 Research approach ... 12 2.5 Scope ... 12 2.6 Operationalization ... 13
3. DIAGNOSIS ...14
3.1 Production process ... 143.1.1 Order process characteristics ... 14
3.1.2 Manufacturing process characteristics ... 16
3.1.2.1 Manufacturing process type ... 16
3.1.2.2 Shop floor configuration ... 18
3.1.3 Conclusion production characteristics ... 19
3.2 Value Stream Mapping ... 20
3.3 Problem description ... 22
3.3.1 External certification ... 22
3.3.2 Rework ... 22
3.3.3 Waiting times in front of the test bed ... 24
3.4 Current working test bed... 24
3.4.1 Layout ... 24
3.4.2 Testing process ... 25
3.4.3 Dispatching ... 26
3.4.4 Capacity ... 27
2
3.5 Production planning ... 28
3.5.1 General production planning ... 28
3.5.2 Planning board ... 30 3.5.2.1 Goal ... 30 3.5.2.2 Accessories ... 30 3.5.2.3 Working ... 32 3.5.2.5 Findings ... 33 3.6 PPC decisions ... 35
3.6.1 Production management framework ... 35
3.6.2 Operational decision functions ... 36
3.6.1.1 Capacity planning/utilization planning ... 37
3.6.1.2 Order acceptation/delivery time setting ... 37
3.6.1.3 Custom order release ... 37
3.6.1.4 Shop order detail planning ... 38
3.6.1.5 Capacity allocation/capacity variation ... 38
3.6.1.6 Shop order release ... 39
3.7 Conclusion diagnosis ... 40
4. THEORETICAL FRAMEWORK ...42
4.1 Pull production ... 42 4.2 Workload control ... 43 4.2.1 Input/output control ... 44 4.2.2 Workload concepts ... 464.2.2.1 Converted load concept ... 46
4.2.2.2 Aggregate load concept ... 47
4.2.2.3 Shop load concept ... 47
4.3 Feedback ... 48
5. DESIGN ...49
5.1 Redesign test bed ... 49
5.1.1 Layout ... 49
5.1.2 Capacity ... 50
5.1.3 Dispatching ... 51
5.2 Input control decisions ... 54
5.2.1 Workload norm determination ... 55
5.2.1.1 Planned output ... 56
3
5.2.1.3 Workload contribution ... 58
5.2.2 Capacity planning ... 60
5.2.3 Order acceptation/due date setting ... 62
5.2.4 Order release ... 67
5.2.5 Order dispatching ... 71
5.3 Output control decisions ... 71
5.3.1 Short-term feedback ... 71
5.3.2 Medium-term feedback ... 73
5.4 The operation of the planning board ... 75
6. CONCLUSION AND RECOMMENDATIONS...78
6.1 Conclusion ... 78
6.2 Recommendations ... 79
4
INTRODUCTION
This thesis presents a case study that is executed at the manufacturing facility of SPX Process Equipment NL BV (further mentioned as SPX NL) in Assen. Spread over six chapters, the research is concerned with the influence of a production, planning and control system on the manufacturing process of the company. The first chapter starts with a brief overview of the company profile, the core activities, the products and the manufacturing process. The second chapter introduces the initial problem statement and defines the research design that is used during the research. It ends with a formulation of the goal and the research questions. The third chapter entails the diagnosis phase. The fourth chapter serves a theoretical foundation for the redesigned PPC system that will be proposed in chapter five. Finally, the sixth chapter provides the conclusions and recommendations of this research.
1.1
SPX Process Equipment
Since December 2006, SPX NL is part of the American company SPX Process Equipment. With a worldwide workforce of 17.000 employees located in more than 35 countries, SPX Process Equipment is a multinational that markets more than 100 brands in four market segments: Flow Technology, Test and Measurement, Thermal Equipment and Industrial Products. SPX NL is the company name of the business unit Netherlands and operates in the Flow Technology segment in which it produces centrifugal pumps. These pumps are sold in a market that covers the regions Europe, the Middle East and Africa and are applied in industries as food, dairy, beverage, biotechnology and pharmaceutics. The business unit Netherlands is divided into two settlements; a main settlement and a service center. The main settlement is located in Assen and houses a production facility. This production facility provides employment for 110 employees, spread over different departments.
1.2
Products
Centrifugal pumps are in general the most used pumps for the transportation of liquids, without adding extra pressure. They transfer low viscosity liquids efficiently within the field of volume and pressure through centrifugal forces. Figure 1.1 represents the cross-section of a centrifugal pump.
5
Figure 1.1 Cross-section of a centrifugal pump
1
2
3 4 5
Centrifugal forces are the result of an impeller (1) that is rotating on an axis (2), pushing a liquid to the outside of a pump shaft (3). Mechanical energy coming from rotation of the impeller is hereby transformed into potential energy (in the form of expanding pressure) and kinetic energy (in the form of increasing velocity). From this, a vacuum originates in the heart of the pump, which drains away new liquid from the suction side, the inlet (4). The extent to which new liquid is drained away from the inlet depends on the rotation velocity of the axis (and the impeller) and the density of the liquid. Finally, the liquid is pushed into the direction of the pressure side, the outlet (5), which in turn converts velocity into height.
6
Figure 1.2 Product structure SPX NL
Combi pumps
Horizontal vertical In-line
Multi-stage pumps DW pumps
vertical In-line Combi Bloc Combi Chem Combi Dirt H Combi Norm Combi Pro Combi Mag Combi Magbloc PHA/HCR Combi Prime H Combi Dirt V Combi Flex Combi Flex Uni
Combi Sump Combi Well Combi Dirt V Combi Line Combi Linebloc MCH MCHZ MCV Centrifugal pumps
The product portfolio can also be classified according to order types. SPX NL distinguishes between standard orders, customer specific orders and project orders. This classification is based on the extent to which an online pump configuration program is able to meet customer requirements. All pumps from the DW group belong to the standard assortment. The Combi group and the Multi-stage group contain a relatively small amount of standard pumps and project orders. These groups include mainly pumps that belong to customer specific orders.
1.3
Manufacturing process
7
Figure 1.3 The production process of SPX NL
Assembling Testing Painting Expediting Mechanical processing 1.3.1 Mechanical department
In the mechanical department non-standard components are manufactured. Raw casting parts are processed en prepared for further operations. Parts that have to meet certain customer requirements, like impellers with a specific diameter, thickness or other non-standard size, are picked from the warehouse and processed until customer requirements are met. Standard components only pass the mechanical department when they have to undergo post-mechanical operations like drilling holes in the pump foot in case a pump has to be connected to a bottom plate. Operations in the mechanical department are drilling, pressing, milling, welding and turning lathe.
1.3.2 Assembling
In the assembly department, pump parts are put together and assembled into a single pump. SPX NL houses four assembly cells; the DW cell, the Combi cell, the MS cell and the traditional cell. In the DW cell only standard pumps are assembled. Pumps from the DW family are responsible for the cooling of ship engines for one single customer. All parts are present in stock and most of them are located in shelves in the assembly cell itself. Some parts have to be picked from the warehouse because their size makes storage on a shelf impossible. Once all necessary parts have been picked, the assembly worker assembles the DW pump according to its work instructions. The customer specific pumps are assembled in the Combi cell, the MS cell and the traditional cell. The most parts are custom made in the mechanical department. Some parts have to be purchased and some common parts are present in the warehouse. The assembly process starts once all parts are available.
1.3.3 Testing
8
Figure 1.4 Schematic representation of the test bed
Tests that can be performed are based on the hydraulic performance and hydrostatic performance of a pump. The most occurring tests are the QHP test and the NPSH test. The QHP test shows the relationship between a volume rate (Q), head (H) and power (P). The NPSH test is used for checking pumps on cavitations, and shows the difference between the pressure and the liquid vapor pressure in a certain cross section of a generic hydraulic circuit1. SPX NL executes six types of tests, called work centers, which differ in the hourly wages that are attributed to the cost price of a pump. Each work center will be described below.
Sample and capacity test. This is a common test that is most frequently demanded by a customer. Both the QHP test and the NPSH test can be performed in this work center.
Endurance test. This test is mostly performed on pumps from project orders and on newly developed pump types by the R&D department. Since no testing information is available, this is a very time consuming test that can take a whole day.
R&D test. This is a work center that is especially designated to R&D. It is mostly performed to check certain components for cavitations, vibrations and noise.
Statistical quality test. This internal work center is aimed at quality control by means of sampling. Sample tests are based on clusters of pumps that are representative for a whole pump group. On average five percent of all manufactured pumps is selected for an internal quality test. If pumps from a certain pump group are very incidentallymanufactured, a rule stating that each pump group has to undergo an internal quality test at least
9
every two years is devised. Some pumps, however, are always tested in this work center, because their design is extremely failure sensitive.
Witness test. Some customers require a second extra test which has to be certified by an external inspector. In this case, a pump first has to pass the sample test before a date with an certification office is planned. On this date the initial test is performed again under the supervision of a certified inspector, who forms a formal test report for the customer.
Calculation and graph drawing. This work center encompasses the additional operations of a finished test. These operations are always performed in the fixed sequence: drawing a graph, checking whether test outcomes are conform a standard norm and closing the test operation. This work center is the final stage of the test bed and is always preceded by one of the previous mentioned work centers.
1.3.4 Painting
10
RESEARCH DESIGN
Any scientific investigation must begin with a structure, or design, which defines the number and type of entities and variables to be studied and their relationship. The major purpose of a design is to answer some specific question based on principles of scientific analysis. This research sets off with the initial problem statement and continues with the formulation of a research question that compasses this problem. The entities subject to this research will be explained by means of a conceptual model. Then, sub-questions that arise from this model will be developed and the methods practiced to answer them will be clarified. Finally the scope of the research will be set.
2.1
Initial problem statement and research question
Since the introduction of a lean manufacturing strategy in 2006, SPX NL has established several changes aimed at the elimination of waste. The implementation of a Kanban system for the standard orders can be seen as the most prominent improvement that was made with regard to the achievement of lead time reduction and just-in-time delivery. In addition, two programs aimed at minimizing material shortages and signaling and minimizing the causes of production disturbances are implemented to control lead times and to warrant just-in-time (JIT) deliveries.
Given that SPX NL faces structural difficulties in being a reliable supplier of centrifugal pumps led in 2008 to a research by Ivo Stikvoort. This research was aimed at determining factors that influence the delivery performance of SPX NL. His major outcomes were inefficient material handlings, the emergence of the test bed as the structural bottleneck station and long waiting times for external inspections. Stikvoort (2008) concluded that inquiring different PPC systems should result in a better delivery performance. This study follows the research of Stikvoort (2008) and supports the management of SPX NL towards a more lean organization. It addresses the current PPC system with the aim of realizing a controlled order stream on the shop floor. Following the management of SPX NL, this research is conducted within the limits of the current capacity constraints. The following research question is the foundation for this research on improving delivery performance:
How can SPX NL control her material and information stream in such a way that delivery performance will increase to 85 percent?
11
2.2
Conceptual model
A conceptual model is a schematic representation of the relations between the entities that ultimately affect a problem. The conceptual model that will be used in this research (figure 2.1) is used to get insight in the acquaintance of the scientific concepts that is requisite to answer the main question.
Figure 2.1 Conceptual model
Production process
characteristics PPC system Delivery performance
It represents the indirect causal relation between variables that determine delivery performance, with production process characteristics as the independent variable, PPC system as the dependent variable and delivery performance as the target variable. Based on scientific literature, it is assumed that the way in which a PPC system affects delivery performance is influenced by production process characteristics. Put differently, a company with certain production characteristics should adopt a suitable PPC system in order to reach a predetermined delivery performance. The relations between the three variables will be explained next.
Production process characteristics/PPC system. According to Altiok (1997) a production system can be seen as an arrangement of tasks and processes, properly put together, to transform a selected group of raw materials and semi-finished products to a set of finished products. Hendry and Kingsman (1989) state that production systems can be characterized according to various factors like product mix, resources, product demand, capacity planning and product lead times. Hopp and Spearman (2000) add that the production process characteristics determine the possibilities and limitations of the manufacturing system with respect to (delivery) performance. Hence, the (re)design of a PPC system is dependent on the production process characteristics.
PPC system/Delivery performance. Following Hendry and Kingsman (2005), a PPC system is needed to meet resources with customer requirements. PPC decisions from different departments guide the order flow towards the customer and feedback information from the order flow is required to monitor progress. This means that PPC systems should facilitate a proper coordination between involved departments in order to meet customer requirements
2.3
Sub-questions
12
NL currently manufactures its products and what shortcomings occur with respect to delivery performance. The design phase is concerned with a redesign of the PPC system that should be able to minimize these shortcomings.
Diagnosis
How can the current production process of SPX NL be characterized? How can the current PPC system of SPX NL be described?
What causes can be identified for the current delivery performance?
Design
How should SPX Process Equipment NL redesign its PPC system in order to improve delivery
performance to 85 percent?
2.4
Research approach
Answers to the research question and the sub questions are obtained by means of an Action Research approach. Benbasat and Goldstein (1987) state that an action researcher is a participant in the implementation of a system, and simultaneously wants to evaluate a certain intervention technique. Checkland and Holwell (1998) add that interpretive action researchers make a clear virtue of the ability of the researcher, and the research process itself, to influence the situation being investigated. The definition of action research by Coughlan and Coghlan (2002) is based on several broad characteristics. First, action research is research in action, rather than research about action. The central idea is that it uses a scientific approach to study the resolution of important social or organizational issues together with those who experience these issues directly. Second, AR is participative, which means that members of the system under investigation participate actively in the process of research. Third, the goal of AR is to make action more effective while simultaneously building up a body of scientific knowledge. Finally, AR is both a sequence of events and an approach to problem solving. Data gathering is facilitated by both active involvement in organizational day-to-day processes and desktop research, and will be performed in different ways. Hard data coming from the manufacturing process, like for instance operational statistics, are gathered from the information system IFS and will be processed in graphs and tables for further analysis. Soft data are generated through literature review, observations, and actual participation in the production process. Feedback on data is given by the researcher’s supervisors and employees that are concerned with the subject of research.
2.5
Scope
13
In addition, the analysis of the order stream will take place on the trajectory of assembly to expedition, because of time constraints. According to Stikvoort (2008) this is the most significant part of the production process when it comes to delivery performance. The mechanical department and the painting department will thus not be taken into account during this research.
This research will come up with a new design of the PPC system. This PPC system is assumed to increase delivery performance, unfortunately time constraints make a quantified analysis of delivery performance impossible. Hence, the implementation phase of the proposed PPC system in the manufacturing process is not included.
2.6
Operationalization
14
DIAGNOSIS
The purpose of the diagnosis phase is to get a clear understanding of how delivery performance is affected by the current PPC system of SPX NL. It starts with the description of the production process characteristics. Then a value stream map is constructed to visualize the current material and information flow on the shop floor. It is aimed at the detection of problems that potentially influence the current delivery performance. After that a general description of the current production planning system is given. A planning board is used as a tool to brighten the current collaboration between the sales department, the planning department and the production department regarding the production planning process. A production management framework is used to get insight in the current interpretation of operational decision functions with respect to production planning and control. Finally, answers to the remaining sub-questions are given and foundations for the design phase are laid.
3.1
Production process
This initial diagnosis describes the current production process of SPX NL and provides an answer to the first sub-question. It starts with a characterization of the order process, based on the position of the customer order decoupling point. Then the manufacturing process is typified in terms of variety and volume and it ends with a description of the shop floor configuration.
3.1.1 Order process characteristics
This section classifies the order process of SPX NL based on the position of the customer order decoupling point (CODP), a term that is defined by Bertrand, Wortmann and Wijngaard (1998) as the point that indicates from where the value-adding material flow is controlled by individual customer orders.
According to the CODP, orders at SPX NL can be characterized as standard, custom or project. Standard orders are concerned with pumps from the standard assortment. Pumps that fall within the standard assortment can be fully composed by the ‘configurator’2, a software program that is used by the sales department for selecting the right components that meet customer requirements. In terms of Hopp and Spearman (2008) the order process of standard orders can be characterized as assemble-to-order (figure 3.1).
2 www.johnson-pump.com/nl
15
Figure 3.1 Assemble-to-order (ATO) process
Assembling Testing Painting Expediting Mechanical processing CODP
The CODP is positioned in front of the assembly operations, which means that all components for standard orders are in stock. A difference is made between components that are purchased, and standard components that have to be pre-processed in the mechanical department. Components that have passed the mechanical department are kept in stock until new orders arrive.
Custom orders have a small degree of standard parts, and unlike pumps from standard orders they are partially composed by the configurator. Pumps that fall within this order classification approximately consist for a small part of standard and purchased components and for a large part of components that are produced in the mechanical processing operation. This order process can be characterized as make-to-order (figure 3.2).
Figure 3.2 Make-to-order (MTO) process
Assembling Testing Painting Expediting Mechanical processing CODP
In the MTO process the CODP is positioned in front of the mechanical department, which means that non-standard components have to be purchased or fabricated after an order is placed. From this moment, all components are bound to a specific order.
16
Figure 3.3 Engineer-to-order (ETO) process
Assembling Testing Painting Expediting Mechanical processing CODP R&D department
Compared to the ATO and MTO processes, the CODP in the ETO process is located further upstream in the production process. As said, the R&D department has to (re)construct components in order to meet specific customer requirements.
Although the order process of SPXC NL can take three different forms, sales figures prove that MTO can be viewed as the dominant process. Since the vast majority of sold pumps stem from customer orders SPX NL is considered as an MTO company in the remaining part of this research.
3.1.2 Manufacturing process characteristics
The manufacturing process characteristics are described according to process type and shop floor configuration.
3.1.2.1 Manufacturing process type
Slack, Chambers and Johnston (2007) state that all manufacturing processes can be sorted according to the dimensions process tasks, process flow, variety and volume. Figure 3.4 represents a continuum of five different manufacturing processes against these dimensions. The dimension process task can vary from diverse and complex to repeated and divided. The dimension process flow is a continuum varying from intermittent to repetitive. Variety implies the amount of standardization within products, with low variety indicating a large amount of standardization and high variety indicating a small amount of standardization. Volume refers to the amount of products being produced at once, i.e. the batch size. In order to characterize the manufacturing process of SPX NL each dimension will be explained in more detail.
17
Figure 3.4 Five manufacturing process types
Project processes Jobbing processes Batch processes Mass processes Continuous processes Volume Variety High Low High Low Process flow Process tasks Intermitten t Repetitive Diverse/ complex Repeated/ divided
Variety. Product variety can be interpreted as high. The product portfolio counts about 1000 different final products, and about 2350 different part numbers3. The products differ in various ways including in their application (water transport, petro chemistry, maritime industry, greenhouses), propulsion (vertical, horizontal, inline, self-suction, immersion), materials (copper, stainless steel, cast iron). Despite the fact that pumps are sold as product modules, in which the interchangeability of parts between pump types is made possible, modularity can be marked as low. The reason for this can be found in the large amount of modules that is currently present.
Volume. A survey of the average batch size during the first half of 2009 shows that the average batch size equals 1,51 with a standard deviation of 1,17. Low batch sizes can be explained by the fact that most customers order only a small amount of pumps. However, in cases in which a customer orders multiple pumps, it may occur that the planning department splits customer orders in multiple shop orders with small batch sizes aimed at WIP reduction. From this can be concluded that volume at SPX NL can be rated as low.
Process tasks. Most of the process tasks are complex due to the high level of specialized skills that is necessary to perform them. The high variety of components and final products requires highly specialized operations in particularly the mechanical department, the assembly department and the testing department. Despite the fact that the low complexity of the painting tasks is independent of the variety of products, the average process task can be rated as complex.
Process flow. Most of the orders follow rather fixed routing steps on department level. Within each department (except by the painting department) however, the product variability is reflected in the variability
18
of operations. This implicates that the amount and type of operations in a department differs between custom orders. Hence, the process flow tends to be moderately intermittent in nature.
From analysis of the four dimension can be concluded the combination of high product variety, small batch sizes, relatively high complex process tasks and a rather intermittent process flow characterizes the manufacturing process of SPX NL as a jobbing process, or job shop. The fact that the functional layout of SPX NL, in which machines are grouped according to their functionality, makes it possible to realize a broad range of orders is another characteristic of a job shop (Van den Berg and Zijm, 1999).
3.1.2.2 Shop floor configuration
In general two different design philosophies with respect to the physical grouping of machines and workers exist (Bokhorst, 2005). In the product-oriented layout, machines and workers are grouped according to manufacturing needs of product types and in the process-oriented layout they are grouped according to the various functions needed to perform all product types.
Figure 3.5 Shop floor layout of SPX NL
Milling pivotting Drilling DW Cell Combi Cell Combi Cell Project Cell Test bed Paint shop Expedition MC Cell
19
characterized by strongly varying product routings in which the order lead time is often a multiple of the order processing times, fluctuating demand for capacity at different operations resulting in continuously varying bottlenecks. In order to respond to this fluctuating demand it is important to have sufficient capacitive flexibility. In the case of SPX NL, however, only some of these characteristics apply. The explanation for this can be found by Land and Gaalman (2004), who make a further distinction between job shops based on the dimensions routing length and routing sequence (table 3.1).
Table 3.1 Shop configurations (Land and Gaalman, 2004)
Routing length
Variable Constant
Routing sequence Undirected Pure job shop Restricted job shop
Directed General flow shop Pure flow shop
The application of these dimensions on the production process of SPX NL points the manufacturing process type towards a general flow shop. The variability in the routing length of jobs is highly dependent on product variability. This relation is confirmed by the fact that complex pumps that are generally composed of many components have more operations (and thus longer routing lengths) than less complex pumps with fewer components. However, the high product and component variability has little impact on the routing sequence variability. With some exceptions, the routings on the shop floor are determined equally, with the visited stations in order of increasing station number. Therefore, the dominant path (as shown in figure 3.5)
mechanical department assembly department test bed paint shop expedition that is pursued by
practically all jobs, supposes the routing sequence to be directed.
The fact that the manufacturing process can be labeled as a general flow shop means that the fluctuating demand for capacity at the work stations mainly stems from the varying production volume of shop orders. Hence, constantly varying bottlenecks as a result of high routing sequence variability do not occur. This is recognized by Stikvoort (2008), who determined the test bed as the only constant bottleneck in the production process of SPX NL. Analysis of this potential bottleneck will be elaborated in the main diagnosis.
3.1.3 Conclusion production characteristics
20
Wijngaard (1990) add that for the control of production in job shops, delivery date setting and lead times become the major points of interest.
3.2
Value Stream Mapping
As production, planning and control address decisions on the acquisition, utilization and allocation of production resources to satisfy customer requirements (Hopp and Spearman, 2008), an analysis of the manufacturing system is needed before making decisions regarding the PPC system. In this section a value stream map (VSM) is used as a first step in analyzing the manufacturing system of SPX NL by modeling the door-to-door material flow.
21
Figure 3.6 Current State Map (CSM) of the door-to-door materials and information flow.
22
The CSM indicates non-value added activities with blasts. All cycle times and queuing times are determined from empirical findings from real time measurement as well as from the analysis of information stored in IFS. On average between 86 and 96 percent of the total throughput time results from non-value added activities. This rather extreme percentages are caused by three main problems, which can be defined as waste resulting from waiting and defects in terms of Hines and Rich (1997):
1. Pumps have to wait for external certification (waiting)
2. About 20 percent of tested pumps face one or more disturbances (defects) 3. Pumps have to wait in the buffer in front of the test bed (waiting)
Each problem will be described in more detail in the next section.
3.3
Problem description
The CSM pointed out that the manufacturing process of SPX NL is confronted with three main problems, which are all located around the test bed. This section gives a more detailed description of each mentioned problem.
3.3.1 External certification
SPX NL sells pumps to customers that require a test certificate for quality reasons. Pumps that obtain such a certificate have to be tested twice, first they are tested on general performance like QHP and NPSH and after that the same test(s) occur in the presence of an external certified test inspector. This test is known as a witness test (section 3.1.2). The time between those two tests is on average 40 hours and during this period pumps wait in a designated area. The reason for this relative long waiting time is that appointments with external inspectors cannot be planned in advance since it is always questionable whether the first test is approved directly. In avoiding unnecessary costs like double traveling costs for inspectors that have to be present twice for one pump, the policy of SPX NL states that appointments with external inspectors can only be made for approved tests. The resulting waiting times can thus be seen as inevitable.
3.3.2 Rework
23
Figure 3.7 Causes of disturbances
Disturbances
Figure 3.7 shows the distribution of causes for pumps that faced disturbances in the period from January 1st 2009 till June 1st 2009. During this period, a total of 1296 pumps were tested. As said before, a disturbance is seen as any form of delay that occurs during testing. This means a disturbance does not necessarily lead to rework, but the occurrence of it interferes with the progress of the testing process. During this period, 20 percent of all tested pumps dealt with one or more disturbances. 51 percent of the detected disturbances were caused by incomplete customer order documents or other (not specified) causes and had minor effects on order progress. The other 49 percent led to rework with notable delays in throughput times as a result. The main causes of rework are:
Deviation from standard curve (60 percent) Defective pump part (40 percent)
Since incomplete customer order documents can be avoided by the planning department, the real rework rate is assumed to be 10 percent of all tested pumps.
24
3.3.3 Waiting times in front of the test bed
On average, pumps have to wait between 20 and 80 hours before they can be tested. These waiting times can be labeled as platform waiting times (Bertrand, Wortmann and Wijngaard, 1990) and they are excluded from waiting times resulting from external certification and rework. The waiting times are supposed to result from the amount of WIP in the buffer in front of the test bed. Therefore, the amount of WIP in front of the test bed, expressed in the cumulative testing time of all regularly waiting pumps, was measured daily during a two month period. The results from figure 3.8 show a highly fluctuating WIP pattern in front of the test bed.
Figure 3.8 Amount of WIP in the buffer
0 10 20 30 40 50 60 mo n we d fri tu e th u mo n w e d fri tu e th u mo n we d fri tu e th u mo n we d fri tu e th u mo n w e d fri
WIP in front of test bed hours
This, combined with findings from Stikvoort´s research (in which the test bed was indicated as the bottleneck station of the production process) requires a closer look on the production planning system with respect to the test bed. The next section first gives a detailed description of the current working of the test bed.
3.4
Current working test bed
This section explains how the test bed functions as a part of the production process of SPX NL. In succession will be looked at the layout, capacity and the testing process.
3.4.1 Layout
25
Figure 3.9 Schematic representation of the test bed
I 1 2 3 4 II V IV III X VI VII VIII IX a b c d e f g .. ..
= Machine (testing spot)
= Couple of pressure gauges = AC junction
Test bed
..
Figure 3.9 gives a schematic representation of the layout and shows that the test bed consists of ten testing spots (from now on machines), each suitable for running tests on various pumps. The machine marked with an ‘X’ for instance, contains a well that makes testing possible for those pumps that need water drainage. Apart from the machines, the output of the test bed is constrained by other factors like pressure gauges, the electricity grid and personnel. Four couples of pressure gauges measure the amount of pressure and suction that arises during testing. A couple can only be used for one test at a particular moment, which means that maximally four pumps can be tested simultaneously. The electricity grid contains seven AC junctions. Again, one junction for one test at a particular moment is applied here. This means that once a pump is connected to the electricity grid by junction ‘a’, this connection is occupied during the whole test. The next pump in the sequence has to be connected to another junction. Currently there are two full time test operators who serve as the most important constraint. All tests on pumps that arrive from the assembly cells are performed by them and each operator is able to run one test a time. An overview of all testing constraints will be given in section 3.4.4.
3.4.2 Testing process
26
Figure 3.10 Five test operations
Preparatory activities Pump connection Actual test Pump disconnection Finalizing activities
Administrative lead time Manufacturing lead time Administrative lead time
Preparatory activities. Before selecting a pump, test operators check a dispatch list to determine which pump should be transferred from the buffer to the test bed, a process that is described in the next section in more detail. Each pump that is waiting in the buffer is attended to a customer order document, a list containing information concerning the test type and whether or not the pump has to be certified.
Pump connection. Before a pump can be installed, it has to be transported to the testing spot. If a pump cannot be lifted manually due to its size, a pulley is used to transfer it. Once arrived at the testing spot, it has to be connected to the pressure gauge through the press and suction junctions. The installation is completed when the pump is adjoined to the electricity grid. Depending on a pump’s size, both transportation and installation can be time consuming activities.
Actual test. The actual test was already expounded in section 1.3.3. Emphasis should be on the fact that a test operator cannot test multiple pumps simultaneously.
Pump disconnection. After a pump is tested, it has to be removed from the testing spot to clear the way for the next pump. The finalized pump is transferred back to the buffer where it will wait for the next operation.
Finalizing activities. After a test is performed, the test results have to be elaborated. In practice this means that test outcomes have to be compared to certain standards that are available in the information system. In the case of rework, the cause of a defect is reported in a database and additional activities are discussed with the master planning. Once test results are conform standards, a test document is realized and the shop order’s testing operation is closed in the routing plan.
3.4.3 Dispatching
27
the whole process is repeated for each next pump that is waiting in the buffer. In practice however, this rule cannot always be followed, as the required voltage and the limited amount of press and suction junctions as mentioned before constitute restrictions to the sequence of pumps to be tested according to the EDD rule. It happens often that a machine, on which the pump closest to its due date has to be tested, is occupied by another pump. As a result, the first job waiting in the queue has to be passed over and the second or even third job is selected for testing.
Figure 3.11 Deviation between actual sequence and sequence according to EDD rule at test bed.
1 2 3 4 5 6 7 Se q u en ce o f jo b s ac co rd in g to E D D r u le 1 3 5 2 4 6 7 = Pump
= Machine (testing spot)
Figure 3.11 clarifies this by showing a difference in the initial sequence (according to the EDD rule) and the realized sequence of pumps (shown by the numbers near the arrows). Jobs 1, 2 and 3 from the buffer have to be tested at the same machine. Operator A selects job 1 for testing, based on the EDD. Operator B cannot select job 2 and job 3, since their machine is occupied. Therefore job 4 would be the first job that can be tested by operator B. As a result, jobs 2 and 3 have to wait. Hence, the initial EDD sequence (1,2,3,4,5,6,7) deviates from the realized sequence (1,4,2,5,3,7,6) resulting in pumps that are in front of schedule and pumps that are behind schedule. A possible explanation can be found in an unbalanced order stream towards the test bed caused by undirected order releases. This means that the planning department currently releases all orders without taking the future machines into account, leading to blocking in front of those machines. The exact way in which jobs are currently released will be described later.
3.4.4 Capacity
28
Table 3.2 Daily capacity constraints at test bed.
Constraint type Present amount
hours
Amount present Hours
Machine 10 80
Test operator 2 16
Pressure gauge 4 32
AC connection 7 56
Table 3.2 summarizes all factors that constrain the capacity of the test bed. The values are based on a working day consisting of eight hours. If for instance exclusively machines would limit capacity, a daily output equivalent to 80 hours could be realized. However, the presence of two machine operators frustrates this, since they limit daily output by a total of 16 hours.
Résumé
A closer view on the test bed reveals that test operators face difficulties in attending test sequences according to the EDD dispatching rule. This might be caused by undirected order releases with respect to the future machines at the test bed. For a better understanding of the current order stream towards the test bed, additional information regarding the production planning system is required.
3.5
Production planning
This section first describes the current production planning system in general terms, by explaining how the planning department plans orders. Then a planning board will be introduced to show the current interaction between the sales department, the planning and the production department and how this interaction affects the performance of the test bed. It ends with a preliminary conclusion about the causes of the waiting times in front of the test bed.
3.5.1 General production planning
29
Figure 3.12 Schematic representation of a push system (Hopp and Spearman, 2008)
Work station ... Work station
Material requirements planning (MRP) is a plan-push system that computes schedules of what should be started into production based on customer demand in which no a priori WIP limit exists. It is used to coordinate orders from within the plant (so-called jobs) and from outside (so-called purchase orders), and the main focus is on scheduling jobs and purchase orders to satisfy material requirements generated by external demand. Indispensible conditions for a proper functioning of MRP are reliable forecasts, standard products and product routings, who serve as a basis for the Master Production Schedule (Bertrand, 2000).
The planning department of SPX NL receives custom (purchase orders) orders from the sales department once a customer has placed an order. Custom orders are transferred to shop orders (jobs) once the routing plan is known. Information is obtained from the lot-sizing rule (LSR) and the planning lead time (PLT). In the LSR custom orders can be split into multiple shop orders if the lot size of an order is too high. With an average batch size of 1,51 the LSR is rarely executed at SPX NL.Each day, the planning department releases shop orders according to their earliest start date. The PLT is used to determine the start time of a job, which is equal to the due date minus the planned lead time. Referring to Land’s (2004) categorization of job release methods, this can be seen as a backward infinite loading (BIL) method. BIL is a controlled release method that aims at reducing the dispersion of lateness through releasing jobs at the right time relative to their due date by estimating the required shop floor throughput time to complete jobs. The release date of a particular job is determined by subtracting its expected process times and queue times from its due date (equation 3.1), where the total queue time functions as a safety lead time used to protect against uncertainties in production timing and demand timing (Hopp and Spearman, 2008).
RD = DD - P - Q (Equation 3.1)
with:
RD = release date
DD = due date
P = total process time
30
Russel and Fry (1997) see BIL as a naive order release approach, since it is a push mechanism that releases orders without explicit regard to timing or shop conditions.
The order progress plan is formed by the routing sheet. This routing sheet captures the routing steps a shop order has to follow, based on customer requirements. Each routing step has its own operation due date which is determined by the delivery date that was set in the previous decision function. The determination of start dates for each routing step is accomplished by putting the routing steps of shop orders in a chronological sequence. All manufacturing steps a job has to pass, have their own planned processing time and planned waiting time, which are estimated from preceding similar orders. Once customer orders are accepted by the sales department, the planning department makes up a routing plan. This routing plan contains information about which operations a shop order has to undergo including the time it takes to perform these operations. For each operation a planned start date is determined based on the planned start date of the next work center minus the planned processing time and the planned queuing time of the current work center. This process is called lead time scheduling by Wiendahl (1995).
3.5.2 Planning board
This section focuses on the consequences of the current production planning system on the performance of the test bed. A planning board (figure 3.13), that was constructed by the author and introduced at the planning department at the first of June 2009, serves as a tool for diagnosing the cause(s) of the highly fluctuating waiting times in front of the test bed.It is supposed to provide a powerful visualization of the consequences of the current MRP system on the amount of WIP in front of the test bed. This section starts with describing the underlying idea of the planning board, then the main features are mentioned together with the way in which additional information is gathered. The design phase shows how the planning board can be used as the communication channel between the three departments.
3.5.2.1 Goal
The use of a planning board is intended to provide insight in how the planning department deals with the amount of accepted orders by the sales department and how this affects the buffer in front of the test bed. Causes of the waiting times are explored through a comparison of the test bed’s daily available capacity with the required capacity resulting from the amount of planned orders. Section 3.4.4 pointed out that the available capacity is constrained by the amount of testing operators, i.e. at a maximum of 16 hours per day. Information regarding the required capacity of a shop order is obtained from the routing plan that is generated by the planning department before an order is released. This routing plan contains information about the expected processing time and start date of a shop order.
3.5.2.2 Accessories
31
different test types. Except for the witness test and the calculation and drawing graph test, all test types have a specific card color. The reason for this is twofold. The witness test can be seen as duplication of the sample test, with the planned start date as the only difference. Planning board cards of witness tests are marked with a ’W’. The calculation and drawing graph test always succeeds another test type and therefore the corresponding processing time of this test type can be added to its predecessor’s card. The total amount of information a card contains is given below.
Figure 3.13 Planning board implemented at the planning department of SPX NL
Mon Tue
Operation due date
Wed Thu Fri Mon Tue Wed Thu Fri Mon Tue Wed Thu Fri
Week 4
Available Required
Week 1 Week 2 Week 3
Time horizon. The planning board has a time span of four weeks. From the first three weeks every single day is shown and the last week is shown as a whole. Available capacity is based on eight hours of effective testing time per operator per day. For each day the total available capacity is determined and written down at the bottom of the plan board. The necessary capacity consists of the total test hours of all planned shop orders during the time span. Every day the necessary capacity is compared to the available capacity and the difference is highlighted.
32
Operation due date (ODD). This is the ultimate date at which a shop order is expected to be processed at the test bed, determined by the lead time scheduling method. Not meeting this date will have serious consequences for delivery performance.
Pump type. The pump type provides information about the supplying assembly cell.
Batch size. The batch size equals the amount of pumps per shop order. Due to low batch sizes, the planning department rarely splits customer orders in multiple shop orders.
Planned processing time. The planned processing time for each pump is determined at the planning department by using the PLT. The total planned processing time of a shop order equals the processing time of a pump multiplied by the amount of pumps.
Released status. Cards of released order are marked with a ‘V’ and the release dates are added.
Execution status. In order to make a visible distinction between executable and non-executable shop orders, planning board cards belonging to shop orders that are waiting in the buffer are marked with a red dot.
Rework status. If a test fails, the card of the corresponding pump is marked with an ‘R’. The pump is transported to the parking area where the cause of failure will be detected. After detection, the pump is re-planned and its card gets a new start date.
3.5.2.3 Working
As said before, the use of the planning board monitors the current way of order acceptation and order release in order to understand how these decisions are related to the amount of WIP in front of the test bed. Information from the planning board was used for a daily comparison between the amount of required capacity (derived from all accepted orders) and the amount of available capacity (derived from the presence of test operators). The next steps were executed by the author on a daily base.
During a period of two months, all accepted orders were captured from IFS4. For each shop order, the previous mentioned information was recorded on a planning board card. All cards were attached to the planning board and ranked according to their ODD within a period of four weeks. For each day the total required processing times were summed up and compared to available capacity. Besides the accepted orders, all released order were captured from IFS5. The planning board cards of released orders were marked with a ‘V’ and supplemented with the release date. Cards of realized orders were subtracted from the planning board and
33
saved in a bin. Cards of rework orders were marked with a ‘P’ and once the cause of rework was detected, a new start date was planned.
Figure 3.14 shows a randomly chosen planning board card of a shop order with Combi Well pumps. Both pumps were released on 15-09-2009 and they required 4,5 hours of processing time at 23-05-2009.
Figure 3.14 Planning board card
22595/23-05-2009 189660 CW 2 19-05-2009 4,5 Test type
Shop order number Batch size Operation start date
Required testing hours Pump type
Release date
V
‘Released’ status
3.5.2.5 Findings
After two months of usage the conclusion can be drawn that the sales department accepts orders and sets delivery dates without taking the capacity of the test bed in to account. The ODD of orders at the test bed are based on the lead time scheduling algorithm, that stems from the routing plan. The planning department releases orders according to the ODD without taking the capacity of the test bed into account. Figure 3.15 shows a comparison of the amount required capacity (indicated by the processing times of accepted orders) and the amount of available capacity (indicated by the presence of two test operators).
Figure 3.15 Required capacity versus available capacity at the test bed (June 1st 2009 – July 31st 2009)
hours 0 5 10 15 20 25 30 35 40 45 mo n w e d fri tu e th u mo n w e d fri tu e th u mo n w e d fri tu e th u mo n w e d fri tu e th u mo n w e d fri
required capacity available capacity
34
The most striking observation is that in some occasions the amount of required capacity exceeds the amount available capacity with a factor 2,5. During the months June and July of 2009 a total amount of 838 processing hours were planned (with a daily average of 19 and a standard deviation of 9,1) whilst only 720 processing hours (equally spread over all days) were available.
During this period, the planning department came up with some actions to lessen the gap between required capacity and available capacity in the short term. In order to decrease the amount of required capacity, some tests were abandoned. In turn, the amount of available capacity was expanded by working overtime.
Abandoning tests. A categorization of testing priorities helped the planning department in determining which orders could pass over the test bed and led to the following priorities:
1. Sample and capacity test 2. Witness test
3. Endurance test 4. R&D test
5. Internal quality test
All internal quality tests (priority 5) and R&D tests (priority 4) were abandoned and all planned endurance tests were postponed. As a result, the total amount of required capacity decreased with 74 hours.
Overtime working. In occasions where critical pumps had to be delivered on time, the test operators worked overtime. During the two months, test operators worked on 12 different days overtime, which resulted in an expansion of available capacity of 44 hours.
Both capacity adjustments resulted in a match between cumulative required capacity cumulative available capacity (table 3.3). However, the highly fluctuating arrival pattern of shop orders at the test bed caused major problems in meeting due dates at this work station. Despite the capacity adjustments, the test bed was not able to meet the ODD for 40 percent of the planned orders. Besides, it should be clear that both capacity adjustments may not be interpreted as structural capacity adjustments. Abandoning tests on a structural base will indefinitely lead to the retraction of quality certificates, which is undesired as quality is highly valued at SPX NL. Structural overtime working might lead to unsatisfied test operators and ultimately to increased probability of sickness.
Table 3.3 Capacity adjustments made by the planning department
Action Required capacity (hrs) Available capacity (hrs)
838 720
Test abandonment -74
Overtime work +44
35
Additional test operators. Adding more test operators to the test bed does not lead to capacity expansions. During a period of two weeks, an extra test operator performed tests at the test bed. The combination of the other constraints (machines, pressure gauges and AC connections) led to an increase in average downtime of all test operators. The output during this period was equal to output in the case of two test operators.
Résumé
The planning board shows the consequences of the current interpretation of the order acceptance and order release decisions for the order throughput at the test bed. The sales department issues unrealistic delivery dates to customers, based on expected lead times of shop orders, without taking capacity of the test bed into account. Orders releases are based on the BIL method, resulting in an unbalanced order stream towards the test bed. Despite of capacity expansions (from abandoning low-priority test and overtime working) test operators are not able to deal with this order stream, leading to major problems in meeting ODD’s and delivery dates in the end.
It is assumed that the current coordination between the sales department, the planning department and the production department results in long waiting times in front of the test bed. The next section addresses the operational decision functions in a PPC system from a literature point of view.
3.6
PPC decisions
The previous section illustrated that the interpretation of the current order acceptation decision and order release decision can be assigned as the cause for the high amount and fluctuating pattern of WIP in front of the test bed. This section inquires the current decisions with respect to production planning and control that are currently taken by the sales, planning and production departments of SPX NL. A production management framework from Bertrand, Wortmann and Wijngaard (1990) will be used to understand the logistic decisions that determine the effect of the current PPC system on the test bed.
3.6.1 Production management framework
