!
!A!new$framework$for!
maintenance(strategy!
development!
In partial fulfillment of the requirements for the degrees of:
MSc in Technology and Operations Management (Rijksuniversiteit Groningen) MSc in Operations and Supply Chain Management (Newcastle University)
Supervisors:
Dr. Jasper Veldman (Rijksuniversiteit Groningen)
Dr. Jingxin Dong (Newcastle University Business School) Thijs Post (Direct Sheet Plant, Tata Steel IJmuiden) Jaap van Dalen (Tata Steel IJmuiden)
Ton Kremers (GWOC)
Abstract!
Purpose – The purpose of this paper is to design a new framework for maintenance strategy
development and to use this framework to analyze the maintenance strategy of the slab guide system of the Direct Sheet Plant (DSP) of Tata Steel IJmuiden and to propose cost-effective improvements.
Methodology – This paper identifies the shortcomings of the currently available maintenance
frameworks and the critical success factors (CSFs) that maintenance frameworks need to achieve. Based on that, a framework developed by Waeyenbergh & Pintelon (2002) is extended. Finally, the new maintenance framework is applied at Tata Steel IJmuiden.
Theoretical contribution – This paper contributes to the knowledge of developing maintenance
strategies by the design of a framework that adds an information management step to facilitate continuous improvement, adds lifecycle management steps to increase quantification and facilitate management authorization, and is flexible and customizable to specific situations while retaining its structure and completeness.
Practical contribution – For the DSP, it was identified that a combination of a roll identification
system and a load monitoring system can improve process reliability, introduce failure prediction, and accomplish significant cost savings.
Value – This paper will be of value to maintenance managers and researchers in the field of
maintenance and asset management. The developed framework is based on published work by other authors; information gathered during a maintenance summer school; and conversations in, experience with and data of the maintenance intensive organization Tata Steel IJmuiden.
Keywords – maintenance framework, maintenance strategy development, maintenance
6.2! OBJECTIVES!AND!SUBESTEPS!OF!INDIVIDUAL!STEPS!...!38! 6.2.1! Step'6:'Design'information'management'...'38! 6.2.2! Step'7:'Asset'lifecycle'evaluation'...'38! 6.2.3! Step'8:'Authorization'and'implementation'...'38! 6.3! CONCLUSION!...!39! 7! APPLICATION'...'41!
7.1! IDENTIFICATION!OF!OBJECTIVES,!KPIS!AND!REQUIRED!PERFORMANCE!...!41!
List!of!acronyms!
BCM: Business-Centered Maintenance
BFID: Bearing Failure Identification Device
bn: Billion
CBM: Condition Based Maintenance
CM: Corrective Maintenance
CSF: Critical Success Factor
DOM: Design Out Maintenance
DSP: Direct Sheet Plant
FMEA: Failure Mode and Effect Analysis
FY: Financial Year
GWOC: GietWalsOnderhoudsCombinatie
KPI: Key Performance Indicator
LCC: Lifecycle Costing
Ltd.: Privately Held Limited Company
MCC: Most Critical Component
MIS: Most Important System
mln: Million
MTBF: Mean Time Between Failures
MTTR: Mean Time To Repair
OEE: Overall Equipment Effectiveness
PM: Preventive Maintenance
RCM: Reliability-Centered Maintenance
RQ: Research Question
SCMD: Strand Condition Monitoring Device
TPM: Total Productive Maintenance
Used!terminology!
Bearing failure identification device (BFID): A device, located at the segment shop floor of
GWOC, that can measure the bearing vibration of all rolls individually without the need to disassemble the rolls from their segment, also called ‘the Wabi’.
Breakthrough: In the first parts of the strand the slab is not yet fully solidified. The slab consists
of a solid skin with liquid steel inside. Different failure mechanisms (e.g. insufficient cooling, heterogeneous skin, impact by the rolls or casting die) can cause the skin to crack. This happens when the solidified skin cannot withstand the pressure arising from the liquid steel (Begoña Santillana 2013) and causes the liquid steel to break through the skin.
Casting distance: The total slab length that is casted in the Direct Sheet Plant at a certain point in
time.
Slab: The steel slab of 70 mm thick that is casted in the Direct Sheet Plant, see figure 2.3.
Slab guide system: A system of six segments that is used to guide the slab from the casting mold
into the tunnel furnace, see figure 2.3.
Static overload: Occurs when a sudden load (mechanical stress) that acts on an asset exceeds the
tensile strength (load-carrying capacity) of the asset (Tinga 2013).
Strand: Term used by Tata Steel to refer to the path (created by the gap of 70 mm between the
top and bottom rolls) that guides the slab from the casting mold into the tunnel furnace.
Strand condition monitoring device (SCMD): Condition monitoring device that can be sent
1 Introduction!
The aim of this paper is twofold. First, this paper proposes a new framework for the development of maintenance strategies. It extends the maintenance framework developed by Waeyenbergh & Pintelon (2002) with the steps ‘design information management’, ‘lifecycle evaluation’, and ‘authorization and implementation’. The second aim of the paper is to analyze the maintenance strategy of the casting section of the Direct Sheet Plant (DSP) of Tata Steel IJmuiden and to propose cost-effective improvements by application of the new framework.
Relevancy!
All man-made structures need maintenance to remain in operation (Dekker, Scarf 1998). This is especially the case for the process industry, which is characterized by capital-intensive industrial plants with huge maintenance expenditures (van Donk, Fransoo 2006). Maintenance directly influences the ability to provide timely services to the customers, and by reducing equipment problems companies can reduce inventories, disruptions, defects and other costs related to malfunctions (Nicholas 1998). Furthermore, new operations strategies, toughening societal expectations, economic pressure, and technological opportunities make that the demands from the maintenance function are more challenging than ever (Tsang 2002). Therefore, maintenance is fundamental to competitive strategy and sustainable performance (Muchiri et al. 2011).
Because of challenging business conditions, Tata Steel is aiming to reduce costs, improve product quality, and improve the reliability of the assets. Keeping the slab guide system of the DSP in good condition is crucial for high productivity and steel quality (Lee, Cho & Kang 1999). However, slab guide system maintenance represents 30% of the direct maintenance expenditures of the DSP and is responsible for high costs of lost production time. Therefore, the DSP wants to analyze the current maintenance strategy of the slab guide system to define cost-effective improvements.
Research!gap!
Theoretical!contribution!
To overcome these shortcomings, a framework as designed by Waeyenbergh & Pintelon (2002) is extended. The new framework aims to continuously improve the maintenance strategy of existing assets based on the objectives of the company. The framework is flexible and customizable by allowing users to use different techniques depending on the situation without losing the overall structure. Furthermore, an ‘information management’ step is introduced to drive proactive data collection, which facilitates optimization and improvement. Finally, ‘lifecycle evaluation’ and ‘authorization and implementation’ steps are introduced to quantify the investment opportunity in economic terms to get management authorization.
Practical!contribution!
Practitioners in asset intensive industries can use the framework for maintenance strategy development and improvement. However, further research is necessary for validation of the framework and extension and adaption of the framework to specific industries. For the DSP, the outcome of the analysis with the maintenance framework could lead to significant performance improvements. A combination of a roll identification and load monitoring system can improve production reliability, introduce failure prediction, and lead to significant cost savings by reducing lost production time. This supports the strategy of Tata Steel in becoming a more cost-effective, reliability focused provider of quality products.
Research!methodology!
This report takes a design science perspective. It is built up as follows; first, the problems of the stakeholders are introduced; second, shortcomings of the currently available maintenance frameworks in literature are identified; third, key steps in the process of developing maintenance strategies are identified; fourth, the critical success factors (CSFs) that the framework and each individual step need to achieve are identified; fifth, based on the CSFs a currently available framework designed by Waeyenbergh & Pintelon (2002) is extended; and sixth, the new maintenance framework is applied at the DSP. The report concludes with an evaluation on the application of the new framework and recommendations for further research.
2 Problem!context!and!problem!definition!
This chapter introduces the stakeholders, the maintenance process, and finally the problem that triggered this research.
2.1 Stakeholders!
There are two stakeholders in this research: Tata Steel and GWOC. This section will introduce the stakeholders to provide relevant background information.
2.1.1 Tata!
Tata is a global enterprise consisting of more than 100 companies spread over seven business sectors. Tata employs over 580.000 people in more than 100 countries. Each Tata company operates independently with its own board of directors and shareholders (Tata group n.d.). One of these companies is Tata Steel. Tata Steel is the world’s sixth largest steel manufacturing company, employs over 80.000 people, and operates in more than 20 countries. Part of Tata Steel is Tata Steel Europe, formerly known as Corus. Tata Steel Europe operates in the UK, the Netherlands, Germany and Belgium (Tata Steel group n.d.). This research focuses on Tata Steel’s biggest site in the Netherlands: Tata Steel IJmuiden. The IJmuiden site employs around 9000 people and produces around seven million tons of steel per year. The IJmuiden site is known for the production of high quality steel, mainly used in the automotive industry, construction and packaging (Tata Steel IJmuiden n.d.).
Though!market!conditions!
figure 2.1: Europe became Tata Steel's main market (adapted from (Malapur 2007)
However, the acquisition of $12,1bn caused Tata Steel’s debt levels to increase vastly (Malapur 2007). Additionally, business suffered under the global financial crisis and subsequently the Eurozone crisis, which had a major impact on the steel markets. In the FY2012-13, the steel demand in Europe was still 30% below pre-crisis levels (Tata Steel Ltd. 2014). In combination with poor performance of the UK sites, the European division kept making losses. In 2014, Tata Steel refinanced its Corus debts (Financial Times 2014), which still puts a lot of pressure on the company’s finances. In combination with increased compliance expenses due to tightening environmental regulations and increased low cost steel production in China, market conditions are especially challenging.
Improvement!program!
As a consequence of the above, in 2010 Tata Steel launched an improvement program that aims to structurally improve the health of the organization through productivity improvement, cost management, optimization of the asset network, and right sizing of manpower (Tata Steel Ltd. 2014). There are three overall goals: improve product quality, reduce costs and improve reliability. Actions to cut costs in the European operations continue in 2014 as conferences are being cancelled, training and IT projects are on hold, and extra levels of authorization on spending are established. “We must continue to manage our costs very aggressively and deliver a step change in results in 2014/15” said Tata Steel Europe CEO Karl Kohler (Tata Steel Insite 2014). This puts pressure on all sites of Tata Steel, including Tata Steel IJmuiden.
! India& 69%& Asia&(ex.& India)& 23%& Rest&of&world& 8%& Europe& 37%& Asia& 24%& UK& 22%& Rest&of&world& 9%& North& America& 8%& Europe& 49%& UK& 29%& North& America& 10%& Asia& 9%& Rest&of&world& 3%&
2.1.2 The!Direct!Sheet!Plant!
At the IJmuiden site, the DSP produces coils of steel from liquid steel in one continuous production line. The DSP started production in April 2000 and has been a project of continuous improvement ever since. During FY2013-14 the DSP reached a record output of 1,26mln tons (Tata Steel Ltd. 2014).
An overview of the DSP is provided in figure 2.2. The first part of the DSP is the thin slab caster: a pan with liquid steel is placed above the continuous casting tank and when the closing of the pan opens, liquid steel will pour into the tank. From the tank the liquid steel is brought into the casting mold, which is built up out of water-cooled copper plates with a spacing that gives the steel slab the right thickness. Cooling makes sure that a solid skin is formed, which increases in thickness down the length of the mold (Begoña Santillana 2013). This partly solidified slab enters the second part of the DSP, the slab guide system. The functions of this system are to support the slab, to transfer the vertical movement of the slab into the horizontal direction, and to extract heat from the slab to fully solidify it. The third part of the system is the tunnel furnace, which makes sure the temperature in the slab is uniform. The fourth part, the mill section, reduces the thickness of the steel to specification. Pressured water cools the slab and removes surface oxides. The final parts of the DSP are the shear and the coiler that respectively cut the slab to length and coils the slab to make it ready for transport.
figure 2.2: Direct Sheet Plant, adapted from Mitsubishi-Hitachi Metals Machinery Inc. (n.d.) Slab!guide!system!
This research focuses on the slab guide system. To obtain high productivity and consistent quality steel, the slab guide system should be kept in good condition (Lee, Cho & Kang 1999). The slab guide system consists of three segment groups (see figure 2.3). The three segment groups have different properties: roll types, angles etc. The main focus in this research is on the rolls of segment group 2-5. In total, there are eight segments 2-5 for four positions in the strand, the
figure 2.3: The slab guide system 2.1.3 GWOC!
All of the maintenance activities of the DSP are fully subcontracted. GietWalsOnderhoudsCombinatie (GWOC) is the maintenance firm that is responsible for the casting section of the DSP. GWOC is a joint venture with Tata steel; Tata steel owns 50% of GWOC’s shares and is the only customer. As a result, GWOC is a highly specialized company. The tasks of GWOC are inspecting and maintaining the casting mold and the segments of the casting section. A big part of the work of GWOC is roll maintenance. To research the condition of the bearings, the rolls used to be disassembled, which is expensive. Therefore, GWOC has invested in a bearing failure identification device (BFID) that measures the bearing vibration. It measures all rolls individually without the need to take the rolls from their frame.
2.2 Current!maintenance!process!
In figure 2.4, the roll maintenance process is depicted. The process has two interwoven closed-loop supply chains, entailing the DSP, transport, and GWOC’s internal departments: segment shop floor, maintenance department, and roll shop floor.
In!strand!inspection!
Once per 72 hours (Broekman 2013) GWOC sends the strand condition-monitoring device (SCMD) through the strand to determine if the installed segments need maintenance. The SCMD measures the roll gap, rotation, alignment and cooling water spray (Lee, Cho & Kang 1999). When the SCMD finds deviations greater than the rejection criteria (Appendix II), the whole segment is exchanged for a spare segment. The failed segment is then transported to the segment shop floor at GWOC.
Segment!shop!floor!
Not all parts of a segment need maintenance. Therefore, the first task for GWOC at the segment shop floor is to check which parts need maintenance. To do this, the segments are first partly disassembled; the top is split from the bottom. Second, the bearing failure identification device (BFID) measures the bearing vibrations. Subsequently, this vibration pattern is sent to the maintenance engineer and is assessed. The rolls with a failed bearing are disassembled and sent to the roll shop floor. The rolls that remained on the segment are measured on the rejection criteria and rejected rolls are also sent to the roll shop floor. The top and bottom part of the segment, including the still installed rolls, remain at the segment shop floor. After that, the top and bottom of the segments are overhauled on not roll related aspects.
Roll!shop!floor!
At the roll shop floor, the failed rolls are again measured on the rejection criteria. Based on these measurements the rolls are overhauled. After that, the rolls are placed in the roll spare stock. Finally, the needed rolls out of the roll spare stock are transported back to the segment shop floor.
Assembly!and!final!measurements!
2.3 Problem+definition+
This section describes the problem as defined by the stakeholders. A problem is regarded as the result of the perception of a state of affairs with which one or more stakeholders are dissatisfied (van Aken, Berends & Van der Bij 2012).
2.3.1 DSP:+High+costs+of+roll+maintenance+
The problem as described by the DSP is that the costs of maintaining the slab guide system are too high. Over 2013, the total maintenance expenditures of the DSP were €23mln, of which €7mln related to slab guide system maintenance. Furthermore, the slab guide system leads to high costs of lost production time. A second problem of the DSP is that the rolls in the production process behave unreliable and roll failures are unpredictable, while the wish of the DSP is to enable scheduled segment exchanges.
2.3.2 GWOC:+The+bearing+failure+identification+device+
GWOC needs to act in accordance with the wishes of the DSP. However, there are some underlying issues that GWOC faces that makes this difficult. GWOC management knows it can cut costs on false rejection of bearings. The maintenance engineer has researched that the deficiencies in the BFID cause that around 40% of the bearings are falsely rejected. Further investments in the BFID could improve this situation. However, GWOC is unsure whether bearing failure identification at the segment shop floor is the right maintenance strategy. This stems from currently running improvement actions. Bearings are being replaced by wider bearings that can handle higher loads, which will improve the lifetime and reliability of the rolls. These improvements could influence the feasibility of investments in the BFID.
The problems of the DSP and GWOC can be formulated into one mutual problem: the need for a cost-effective maintenance strategy for the rolls of the continuous casting machine of the
2.4 Problem+Validation+
Before starting the research, it needs to be investigated whether the described situation is a real-life problem (van Aken, Berends & Van der Bij 2012). In figure 2.5, a failure mode, effect and criticality analysis that the DSP uses to assess the criticality of situations, is shown. The current maintenance process led to an average of one segment (2-5) exchange every 10 days in 2013. Furthermore, this led to an average of 3,5 hours of unplanned downtime per exchange and average direct maintenance costs of €35.000 per segment. Together this results in a situation of which the criticality needs to be reduced. Therefore, this problem is classified as a real-world problem for Tata Steel that is therefore worth researching. Furthermore, segments 2-5 are the segments with the highest cost contribution (figure 2.6). Therefore, the initial focus on segments 2-5 seems justified.
figure 2.5: Failure mode, effect and criticality analysis
figure 2.6: Direct maintenance costs of segment groups 1-6 (over 2013)
4 (critical
effect)
> €100.000 > 7 hrs L < 1 yr L < 1 yr H < 1 wk ZH < 24 hr ZH < 24 hr
3 €10.000 to €100.000 1 tot 7 hrs L < 1 yr L < 1 yr L < 2 month H < 1 wk ZH < 24 hr 2 €1.000 to €10.000 4 mins to 1 hr No action L < 1 yr L < 2 month H < 1 wk ZH < 24 hr 1
(small effect)
3 Research+methodology+
Because this research deals with solving a real world improvement problem, an appropriate methodology is design research (Karlsson 2010). Design research aims to solve real world organizational problems through the use of scientific knowledge and local facts (van Aken, Berends & Van der Bij 2012). Design researchers use combinations of explicit and tacit knowledge developed through experiential learning in a creative process of reflection-on-action, often using a participant-observer-perspective (van Aken, Romme 2012). Different research techniques may be utilized depending on the specific situation (Karlsson 2010). Research techniques used in this research are literature study, data analysis, meetings, observation and participative research. The data that is used in this section is collected from internal databases, Excel files, hardcopy documents, other software, and tacit knowledge of engineers of Tata Steel and GWOC.
Designing is the process of determining the required function of a system, the ways of how a system has to achieve this function, and the development of a system that accomplishes this (van Aken, Berends & Van der Bij 2012). For ‘ways of how the system should achieve its function’, this research uses the term Critical Success Factors (CSFs). CSFs are defined as particular characteristics, conditions, or variables that are critical for the success of a system and therefore should be properly sustained, maintained, or managed (Leidecker, Bruno 1984). Therefore, before a proper system can be designed its CSFs have to be identified.
table 3.1: Research methodology overview
Phase 1 (Problem analysis): section 4
• The initial problem is analyzed to set the scope and provide direction to the research.
Key factors that the new maintenance framework needs to contain to solve the problem of Tata Steel are identified.
• Based on the identified key factors, research questions are determined. They are
divided over two main objectives: To make a contribution to the knowledge of developing maintenance strategies and to propose improvements to the maintenance strategy of the DSP.
Phase 2 (Literature review): section 5
• A literature review is conducted to answer the theory research questions. It aims to
answer what characteristics maintenance frameworks need to have, what the most important steps in maintenance frameworks are, and how information management and lifecycle management can be incorporated in a maintenance framework.
Phase 3 (Design): section 6
• The knowledge from the theoretical research questions is synthesized into a generic
model for maintenance strategy development. Furthermore, sub-steps and techniques are identified in order to be able to put the framework in its applications context.
Phase 4 (Application): section 7
• The new framework, together with the identified sub-steps and techniques, is applied to
provide improvement directions to the maintenance strategy of the casting section of the DSP.
Phase 5 (Evaluation): sections 7.8 and 8
• Section 7.8 (Conclusion) evaluates the identified improvements to the maintenance
strategy of the casting section of the DSP and looks for next steps to undertake for the DSP.
• Section 8 (Discussion and Conclusion) evaluates the extent to which the new
maintenance framework meets the desired specifications, addresses its limitations and gives recommendations for further research.
4 Problem+analysis+
In this section the problem will be analyzed to identify the problems. Based on the sub-problems, a set of research questions (RQs) is formulated.
4.1 Analysis+of+defined+problem+
Four sub-problems with the current maintenance strategy have been identified (figure 4.1). Together, these problems explain why the defined problem is hard to solve. These sub-problems are divided into key factors that should form an integral part of the maintenance strategy.
Section 2.1 explained that the rolls have no identification number and hence, are not being traced through the closed loop maintenance supply chains (see figure 2.4). This makes it hard to link rolls to specific failure data, specific positions in the process and makes it impossible to analyze the lifetime distribution of the rolls. Second, a lot of experiments are being conducted with rolls. This causes that different types of rolls (e.g. normal rolls and rolls with widened bearings), with different characteristics, are existent in the process, making the data hard to analyze. The third problem is that weak IT integration between the DSP and GWOC cause that there is only a weak insight in the influence of process behavior on the condition of specific rolls. For example, breakthroughs are expected to have a big influence on the failures of rolls; the effects however are not well documented. The fourth problem is that the
improvement developments have a yet unknown influence on the feasibility of the
maintenance strategy. For example, widened bearings are expected to prolong the lifetime of the rolls, which has a yet unknown influence on the feasibility of investments in the BFID. Together, the four identified sub-problems make it hard to define a cost-effective maintenance strategy.
figure 4.1 Conceptual model of problem analysis
!
Informa(on) Management) Problem)with)IT) integra(on)Tata) and)GWOC) Weak)insight)in) rela(on)to)process) behavior) Roll)) experiments) Weak)insight)in)roll) failure)behavior) Roll)) iden(fica(on) Hard%to%design% Cost.effec1ve% maintenance% strategy% Unknown) influence)of) developments)on) strategy) Lifecycle) Management) Unknown) influence)of)future) developments)Iden1fied%sub.problems% Effect% Key%factors%in%maintenance%
4.2 Research+questions+
This research has two objectives. The first objective is to make a contribution to the knowledge of developing maintenance strategies. The second objective is to propose improvements to the maintenance strategy of the DSP. Both objectives have separate RQs (table 4.1).
The theory research consists of answering the first three RQs. In order to propose improvements to the maintenance strategy of the DSP, a frameworks needs to be used that is suited to the problem situation. Therefore, RQ1-3 aim to identify the characteristics that this maintenance framework needs to have. To do this, it is researched what the most important steps are, what their problems are and how information management and lifecycle management could be incorporated in the framework. RQ4 aims to synthesize the knowledge gained into a generic model for maintenance strategy development and identifies sub-steps and techniques to be able to put the framework into practice. Finally, RQ5 aims to find improvement directions for the maintenance strategy of the casting section of the DSP utilizing the new maintenance framework.
table 4.1: Research questions
Literature 1. What are the characteristics that maintenance frameworks need to have? 2. What are the most important steps in maintenance frameworks and what are
their objectives and problems?
3. How can information management and lifecycle management be incorporated into a maintenance framework?
Practice 4. What framework can be used to improve the maintenance strategy of the casting section of the DSP?
5. How can the maintenance strategy of the casting section of the DSP be improved?
5 Literature+review+
This section answers the theory RQs. To do this, this section is divided according to the first three RQs from table 4.1. Each individual RQ is answered in the conclusion section. Some definitions are explained first. The overall goal of maintenance is total asset lifecycle optimization (Pintelon, Parodi-Herz 2008), which means optimizing availability and reliability of assets to be able to produce the right quantity and quality of products on time against the lowest costs over the entire lifecycle. Maintenance strategy is a plan of action to achieve this long-term overall aim (OED 2014). In particular, maintenance strategies determine which maintenance policies and at what time are the most suitable in reaction to certain failures (Dekker, Scarf 1998). A maintenance framework is a method, or decision support tool, to define a maintenance strategy for a specific situation (Waeyenbergh, Pintelon 2004). A maintenance framework is used to plan, control and improve the different maintenance policies applied in a company. A typology of maintenance policies is provided in Appendix III.
5.1 What+are+the+characteristics+that+maintenance+frameworks+need+to+have?+
Literature describes a number of maintenance frameworks. The first generation is the maintenance policy decision tree with specific yes-or-no questions, leading to certain maintenance policies (Pintelon, Parodi-Herz 2008). Decision trees are still used as part of maintenance frameworks.
When assets became more complex systems there was a need for more integrated maintenance frameworks (Pintelon, Parodi-Herz 2008). The first example is lifecycle costing (LCC), which emphasizes total cost of ownership principles. The asset lifecycle management model (Schuman, Brent 2005) is an example of an application of LCC. The second example is total productive maintenance (TPM), which is a Japanese maintenance philosophy similar to total quality management and LEAN. Like LEAN, TPM focuses on increasing effectiveness by eliminating waste and shares much of its tools, such as; single minute exchange of die, poke yoke, jidoka etc. Furthermore, it puts emphasis on the measure overall equipment effectiveness (OEE), see section 5.2.1.
framework (Braaksma, Klingenberg & Veldman 2013, Rausand 1998). In RCM, a failure mode and effect analysis (FMEA) of critical items is made, based on which certain maintenance policies are applied. The final example is business-centered maintenance (BCM), a framework that builds on RCM but puts the business central (Pintelon, Parodi-Herz 2008). It aims to establish a maintenance strategy that fits the business objectives.
5.1.1 Downsides+and+upsides+of+traditional+maintenance+frameworks+
Waeyenbergh & Pintelon (2002) argued that many companies have difficulties implementing the described frameworks because companies need a framework that is customizable instead of a standard solution. In table 5.1, the most important downsides of the traditional frameworks are summarized. The main drawback of RCM is that it focuses on reliability instead of economics, this might be suitable for the airline industry but not for the process industry (Waeyenbergh, Pintelon 2002). Furthermore, both RCM and BCM are very complex and rely on an extensive amount of data, data that is often not available (Braaksma 2012). TPM is a philosophy for continuous improvement, but gives no guidelines for which maintenance policy to choose. The same applies to LCC, as it provides no clear procedure.
table 5.1: Downsides of maintenance frameworks, adapted from Waeyenbergh & Pintelon (2002)
RCM BCM TPM LCC
Complex, extensive need of data, focus on reliability, no feedback Complex, extensive need of data No maintenance framework but philosophy, no costs taken into account
No maintenance framework but philosophy, no standard procedure
Waeyenbergh & Pintelon (2002) also state positive aspects of the traditional maintenance frameworks. RCM is a structured method that relies on rational decisions leading to traceable outcomes. BCM is business centered. LCC focuses on the integration of the entire asset lifecycle. Additionally, they concluded that an effective framework considers maintenance holistically; it should take into account both the technical asset characteristics and the relation with the production process. Finally, they stress the need that the framework can be fully tailored to specific situations. These important aspects are summarized in table 5.2.
table 5.2: Upsides of maintenance frameworks, adapted from Waeyenbergh & Pintelon (2002)
• Structured and rational with traceable outcomes
Based on the identified upsides and downsides, Waeyenbergh & Pintelon (2002) developed a new framework (figure 5.1). They mention that in future research, the framework should be further extended with aspects such as data management, replacement policy, lifecycle costing and standardization and that each phase should be carefully elaborated.
figure 5.1: Maintenance framework as developed by Waeyenbergh & Pintelon (2002)
5.2 What+are+the+most+important+steps+in+maintenance+frameworks+and+what+
are+their+objectives+and+problems?+
This section elaborates on the steps as described by Waeyenbergh & Pintelon (2002) to identify their objectives, sub-steps, and pitfalls to research improvement and extension opportunities.
5.2.1 Identification+of+objectives+and+resources+
Maintenance strategy should support the corporate strategy (Salonen 2009). Therefore, maintenance strategy definition requires the maintenance objectives (see figure 5.2) and KPIs (Uday et al. 2009). This enables to judge the effectiveness of the maintenance strategy in supporting the corporate strategy. When a maintenance strategy is not aligned with the objectives, a new one needs to be developed. Whether this is the case, can be investigated using three important indicators: inherent, required, and actual performance of the asset (Waeyenbergh, Pintelon 2004).
figure 5.2: Maintenance objectives, adapted from (Muchiri et al. 2011) 1:#Iden(fica(on#of#objec(ves#
and#resources# 2:#Iden(fica(on#of#the#most#important#systems# 3:#Cri(cality#analysis#
4:#Maintenance#policy# decision#step# 5:#Op(miza(on#of# maintenance#policy# 6:#Performance# measurement#and# con(nuous#improvement#
Corporate(strategy( Manufacturing(strategy(
Muchiri et al. (2011) give an overview of the most important maintenance KPIs in the OEE framework, see figure 5.3. The authors state that recommended targets are: preventive maintenance (PM) 75%-80% and reactive maintenance (RM) 10%-15%. They also advise that 5%-10% of the hours should go to improvement work. High percentages of reactive maintenance can lead to quality problems and high costs of downtime. Therefore, the focus should be on continuous improvement of equipment to reduce the need for maintenance (Pinjala, Pintelon & Vereecke 2006).
figure 5.3: Important maintenance KPIs in OEE framework, adapted from (Muchiri et al. 2011) 5.2.2 Identification+of+the+most+important+systems+(MIS)+
The goal of this phase is to focus the work on the most important system (MIS), which decreases the resource intensiveness that was found to be a problem of the traditional approaches. The MIS often shows a high frequency of functional failure, defined as the inability of an asset to fulfill a function to its performance standard (Moubray 1997). The probability that a piece or component will perform according to its performance standard over a specific time period is called reliability. The lower the equipment reliability, the more often the need for maintenance (Dhillon 2002). Information that can be used in this phase is process data, survival functions, financial information, experience from maintenance personnel etc.
5.2.3 Identification+of+the+most+critical+components+(MCC)+
To implement cost-effective maintenance policies one needs to know which functional failures have consequences that have to be prevented (Moubray 1997). It is impossible to prevent all failures; therefore the focus must be on the most critical components (MCC). The MCC within the MIS can be identified using a simplified failure mode and effect analysis (FMEA) (Waeyenbergh, Pintelon 2002). FMEA is a procedure to identify failure modes including their probability and consequences. However, Braaksma, Klingenberg & Veldman (2013) show fundamental problems with the FMEA procedure, especially related to information
Failure+modes+
A failure mode is defined as an event that causes a functional failure (Moubray 1997). Suzaki (1987) divided causes of equipment failures into five categories. First, any equipment deteriorates or wears out. Second, equipment that is not designed for the purpose it is used for wears out faster. Third, bad maintenance (e.g. lack of cleaning and lubrication) will result in accelerated deterioration. Fourth, tough operating conditions will accelerate deterioration. Fifth, bad operation of the equipment can also result in accelerated deterioration. Moubray (1997) states that the load carrying capacity of an asset must always be higher than the ‘worst case’ external load.
Failure+effects+
For cost-effective maintenance there is a trade-off between letting equipment fail and performing maintenance. To make this trade-off, the effects of a failure need to be studied. Nakajima (1988) divided these effects into three groups: availability problems, downtime from equipment setups and chronic equipment breakdowns; efficiency problems, idling, caused by jams and reduced speed and yield; and quality problems, nonconforming output caused by variability in the process. All of these effects influence operating cost, customer order fulfillment, and productivity.
5.2.4 Maintenance+policy+decision+
After selecting the MCC, one needs to make a decision which maintenance policy to apply. Numerous maintenance policy decision trees have been developed in literature (Moubray 1997, Waeyenbergh, Pintelon 2002, Waeyenbergh, Pintelon 2004). This section will summarize the basic deliberations behind maintenance policy decisions.
Design out maintenance (DOM) is often the preferred maintenance policy since it can reduce or eliminate the need for maintenance. If DOM is infeasible (technically of economically) condition based maintenance (CBM) is the next preferred policy since it enables right on time maintenance actions. However, CBM needs data that often comes from sensors on the asset (failure data) or in the process (process data) and models with which the failure can be predicted (Veldman, Wortmann & Klingenberg 2011), which is not always feasible.
maintenance). In practice however, age related failure patterns are not often found since failures are often related to static overloads (Tinga 2010). Furthermore, in an uncertain and changing environment it is very difficult to plan maintenance activities based on historic data (Yam et al. 2001). When PM and CBM are not feasible, corrective maintenance (CM) is an option. This often means that redundancy has to be installed to be able to replace the failed asset when necessary. Furthermore, when the failure is hidden, a scheduled failure-finding task is needed (Moubray 1997).
5.2.5 Optimization+of+maintenance+policy+
This phase is about the optimization of the chosen maintenance policies. For PM there are two classic options: age-based maintenance and block-based maintenance (Waeyenbergh, Pintelon 2002). During the optimization phase, different maintenance policies can be combined to reach optimal effectiveness. Although CBM is increasingly applied, especially in the process industries, little guidance is available for selecting a certain type of CBM (Veldman, Wortmann & Klingenberg 2011).
5.2.6 Performance+measurement+and+continuous+improvement+
Industrial systems evolve rapidly which makes that the maintenance frameworks need to take into account the changing systems and changing environments (Waeyenbergh, Pintelon 2002). In order to facilitate continuous improvement, performance measurement and reporting is crucial because it can identify gaps between current and desired performance, which can help maintenance managers to focus resources (Muchiri et al. 2011). However, maintenance performance reporting is difficult because of a time-lag effect between maintenance execution and plant performance and because of that plant performance is influenced by much more than the maintenance function (Waeyenbergh, Pintelon 2002).
5.3 How+ can+ information+ management+ and+ lifecycle+ management+ be+
incorporated?+
In section 4.1, information management and lifecycle management, were identified as important in solving the problem of defining a cost-effective maintenance strategy for Tata Steel. However, these factors are missing in existing maintenance frameworks.
5.3.1 Information+management+
Information management is crucial, the more failure data is collected the more precise the predictions of failures can be established (Veldman, Wortmann & Klingenberg 2011). Maintenance frameworks often assume that knowledge is readily accessible by all members of the maintenance staff when often the opposite is true (Naughton, Tiernan 2012). Several authors have identified information management as an origin of problems in the maintenance environment (Braaksma 2012). Braaksma (2012) has grouped these into seven core problems (table 5.3).
table 5.3: Information management problems, adapted from (Braaksma 2012)
1. Uncertainty of future information needs: It is unclear which data needs to be collected for future asset management.
2. Maintaining high quality asset data is costly and complex: The potential value of asset information is not known.
3. Maintenance knowledge is insufficiently accessible: Much of the information is embodied in a person.
4. Information cannot be used without additional knowledge: Asset data is stored without sufficient context to be used effectively.
5. Heterogeneity of storage applications: Data is stored in several non-integrated systems: This complicates analysis.
6. Data hand-over problems: Problems of handing over data between the stakeholders of different asset lifecycle phases.
7. Lack of information standards: Complicates the exchange of asset data.
on paper, making structuring and analyzing difficult (Waeyenbergh, Pintelon 2002). Analyzing is especially difficult because of the fourth problem; data cannot be used without additional knowledge because asset data is stored without context. The fifth problem is the heterogeneity of storage systems. Maintenance information systems are often non-integrated and designed and used only for storing information instead for retrieving and analyzing information (Garg, Deshmukh 2006). The sixth problem is the data hand-over between stakeholders. There often exists a gap between process and maintenance engineering (Veldman, Wortmann & Klingenberg 2011), which makes that process information is often not available (Naughton, Tiernan 2012) leading to incomplete maintenance information. Finally, the seventh problem is the use of several information standards, making it hard to combine data from several non-integrated sources.
5.3.2 Lifecycle+management+
5.4 Conclusion+
The aim of the literature review was to identify the characteristics that a maintenance framework, in order to solve the problem of Tata Steel, needs to have. This section answers the three corresponding RQs.
RQ1: What are the characteristics that maintenance frameworks need to have? Maintenance
frameworks should be structured and complete so that all aspects are taken into account, yet providing customizability (Waeyenbergh, Pintelon 2002). The traditional frameworks (RCM, BCM, TPM, LCC) are often not applicable in practice (Naughton, Tiernan 2012, Waeyenbergh, Pintelon 2002). The available maintenance frameworks are often found to be too resource intensive, not flexible, not customizable, not suitable for continuous improvement and reliant on non-existing information (Braaksma 2012). The strong points of the traditional methods are: structured approach, business centered, taking the whole organization into account, performance improvement and cost savings. These weak and strong points are brought together in a set of CSFs that a maintenance framework needs to fulfill (table 5.4).
table 5.4: Critical success factors of maintenance frameworks
• Enable continuous improvement
• Based on business objectives
• Structured and complete procedure with traceable outcomes
• Flexible and able to be tailored to specific situations
• Not only focus on the technical characteristics of the system but the whole organization
• Take into account information collection and management
RQ2: What are the most important steps in maintenance frameworks and what are their objectives and problems? Waeyenbergh and Pintelon (2002) developed a new maintenance
framework consisting of six steps, of which the objectives are stated in table 5.5.
A number of problems with this framework were identified. First, although information management and lifecycle management were found to be of key importance to maintenance strategy, they are not part of any existing framework. Second, FMEA, regarded as an important step of maintenance frameworks, is not suited for continuous improvement (Braaksma, Klingenberg & Veldman 2013). Third, although CBM is increasingly applied, especially in the process industries, little guidance is available for selecting a certain type of CBM (Veldman, Wortmann & Klingenberg 2011). Fourth, maintenance performance reporting is difficult because of a time lag effect between maintenance execution and actual plant performance and because of that plant performance is influenced by much more than the maintenance function (Waeyenbergh, Pintelon 2002).
table 5.5: Steps and objectives in the framework of Waeyenbergh and Pintelon (2002)
Step: Objectives:
Identification of objectives and resources
Identify the objectives and resources of the business and system
Identification of MIS Identify the sub-system that prohibits the total system from reaching the objectives
Identification of MCC Identify the components that cause this sub-system to not reach its functional goal
Maintenance policy decision Select the combination of maintenance policies that fits the sub-system best
Optimization of chosen maintenance policy
Optimize the chosen maintenance policy
RQ3: How can information management and lifecycle management be incorporated in a maintenance framework? Information management and lifecycle management were identified
as important topics in defining a maintenance strategy for Tata Steel.
Information management allows for increased quantification, less reliance on expert judgment, and continuous improvement. Braaksma (2012) proposed four CSFs for maintenance information management (table 5.6). First, data collection should be done proactively. This means that data collection should be organized before the data is needed for analysis. Second, data collection should be worthwhile to do. It is not necessary to collect the data for any single component in the process; instead the focus should be on the most critical components. Third, data collection should be supported by a clear business case. This enables focused communication to operators and increases the accuracy of data collection. Fourth, contextual data, in which data purpose and relevant situational factors are explained, should be stored.
table 5.6: Critical success factors of information management (Braaksma 2012)
1. Proactive data collection
2. Data collection should be worthwhile 3. Data collection supported by business case
4. Storage of contextual data
Asset lifecycle management can provide quantification of investment opportunities in monetary terms to get management approval. Asset lifecycle management offers important benefits as visibility of upstream and downstream costs, analysis of business function interrelationships, and revenue prediction, see table 5.7.
table 5.7: Critical success factors of lifecycle management
1. Quantify upstream and downstream lifecycle costs 2. Quantify business function interrelationships
3. Make revenue prediction
6 Design+
This section answers RQ4: What framework can be used to improve the maintenance strategy
of the casting section of the DSP? To do this, the framework of Waeyenbergh & Pintelon
(2002) will be extended based on the identified CSFs. In particular, information management (step 6) and lifecycle management (steps 7 and 8) are added.
6.1 Design+of+the+new+maintenance+framework+
The first step in designing the new maintenance framework is the identification of the functional goal. The functional goal of the maintenance framework follows from the definition of maintenance (see section 5): To provide a framework to define, control and improve a
maintenance strategy for a specific system so that it achieves the defined objectives over the entire lifecycle of the system. Based on the identified CSFs in table 5.4, a new maintenance
framework is designed, see figure 6.1. How the framework works will be explained first, how the framework fulfills the CSFs will be explained second.
figure 6.1: New maintenance framework
In table 5.4, the CSFs of maintenance frameworks were identified. To enable the CSF
continuous improvement the framework is based on the plan-do-check-act cycle. The cycle
integrates the organization’s objectives with the objectives of lower functions, and subsequently enhances these lower functions to ultimately improve the organization’s overall performance (Qing-Ling et al. 2008). This also fulfills the second CSF, as the whole process is
based on the business objectives. The framework is structured and complete because it is
based on prior research that takes the advantages of the traditional maintenance frameworks together. Still, each phase allows users to utilize different techniques depending on the situation without losing the overall aim. This makes the framework flexible and able to be
tailored to specific situations. Furthermore, compared to the framework of Waeyenbergh &
Pintelon (2002) it adds the phases ‘design information management’, ‘lifecycle evaluation’ and ‘authorization and implementation’ in order to take into account information collection
and management and quantify improvement opportunities for management authorization.
+
2:#Iden(fy#most#important# system# 3:#Iden(fy#most#cri(cal# components# 5:#Op(miza(on#of#6.2 Objectives+and+subVsteps+of+individual+steps+
In table 5.5, the objectives of the different steps in the framework were identified. Each step can be regarded as a framework on its own. Steps 1–5 and 9 were already discussed extensively. This section explains steps 6 to 8, as these are new to the maintenance framework.
6.2.1 Step+6:+Design+information+management+
Based the CSFs of table 5.6, Braaksma (2012) proposed three steps to design a maintenance information system. A slightly adapted version is described in table 6.1. These steps will be followed during application of the new maintenance framework.
table 6.1: Information management system design, adapted from (Braaksma 2012)
Step Corresponding questions
a. Determine if maintenance information collection is worthwhile
• Describe information shortage
• Determine if there is improvement potential
• Determine if improvements are worthwhile
b. Determine data analysis requirements
• Determine what data analysis is needed in the future
• Determine the data requirements for this analysis
c. Prepare data collection • Determine additional actions
6.2.2 Step+7:+Asset+lifecycle+evaluation+
In table 5.7, three CSFs were identified for asset lifecycle evaluation. Schuman & Brent (2005) propose an asset lifecycle management model for assets in the process industry of which the steps can be followed during application of the framework. The first step is to quantify the lifecycle costs. Gram & Schroeder (2012) show the cost factors that can be taken along per lifecycle for the process industry. The second step is to incorporate business function interrelationships. For example, the influence of the investment on the identified KPIs during step 1 of the new maintenance framework can be quantified. The third step should focus on prediction of the revenue that the investment can bring. Ultimately, this step should show the value of the investment opportunity.
6.2.3 Step+8:+Authorization+and+implementation+
6.3 Conclusion+
This section answered RQ4: What framework can be used to improve the maintenance
strategy of the casting section of the DSP? The framework of Waeyenbergh & Pintelon (2002)
figure 6.2: New maintenance framework including followed sub-steps 2:#Iden(fica(on#of#most# important#systems# 3:#Iden(fica(on#of#most# cri(cal#components# 5:#Op(miza(on#of#
7 Application+
This section answers RQ5: How can the maintenance strategy of the casting section of the
DSP be improved? To do this, the new maintenance framework (see figure 6.2) will be
applied. Step 8: authorization and implementation and step 9: performance measurement, are skipped because of time and resource constraints.
7.1 Identification+of+objectives,+KPIs+and+required+performance+
The first step in the new maintenance framework is built up out of three sub-steps: identify objectives, identify corresponding KPIs, and identify required performance.
Tata Steel describes its mission as follows: “To be the long-term preferred partner in our chosen markets by unlocking the potential of steel” (Tata Steel Ltd. 2014). The business objective for Tata Steel IJmuiden is “to make a big contribution to the local and national economy, knowledge, and employment with a strong focus on sustainable business continuity” (Tata Steel IJmuiden n.d.). As stated in section 2.1, in order to ensure this business continuity, Tata Steel has started an improvement program with three objectives; reduce costs of raw materials, maintenance, energy, and personnel; improve quality in a customer focused way; and improve reliability and availability of assets. Furthermore, Tata Steel has a strong focus on quality, product innovation and operational excellence. The mission of GWOC is to maintain all installations of the casting section of the DSP. Therefore, the objective of GWOC conforms to the objectives of Tata Steel.
For the identification of the right objectives and KPIs with respect to this project, the framework of Muchiri et al. (2011) is used. First, the overall objective of the project is to
reduce costs. These costs are divided into two groups: ‘direct maintenance cost’ is the cost of
Required+performance+
No clear performance requirement is set for the objectives and KPIs, however some rough performance requirements can be stated. First, the maintenance manager of the casting section of the DSP wants to reduce the overall costs of segment maintenance with at least 20%. Second, predictability of segment failures is an important objective for Tata Steel. In order to plan segment exchanges, failure predictability should be high. Furthermore, Pinjala, Pintelon & Vereecke (2006) state that the percentage of proactive maintenance should be high in order to be able to produce quality products and reduce maintenance costs. According to Muchiri et al. (2011), a proportion of proactive versus reactive maintenance, appropriate to this objective, would be 85% and 15% respectively. Finally, reliability should be high. Appendix IV shows that in order to be able to plan segment exchanges in the maintenance intervals without running out of spare segments, a minimum MTBF of 16 weeks is needed, corresponding to a failure frequency of 1/544 km. Furthermore, a benchmark with the Oxygen Steel Factory (OSF) shows that the MTBF of segments in the OSF is 68 weeks (Appendix V). This makes a MTBF of 16 weeks in the DSP not an unrealistic goal.
table 7.1: Objectives and KPIs
KPI Description Required
Cost >20% reduction
Direct maintenance cost Lost production cost
Costs of maintenance actions Costs of lost production time
- -
Predictability High
Proactive maintenance Reactive maintenance
Percentage of proactive segment exchanges Percentage of reactive segment exchanges
85% 15%
Reliability High
Failure frequency MTBF
Failures over a certain casting distance Mean time between failures
> 1/544 km > 16 weeks
+
+
7.2 Identification+of+most+important+system+
Predictability+
The average failure frequency of a segment is 1/188 km (MTBF of 5,5 weeks). 91% of the segments are currently exchanged on CM. The DSP’s strategic focus on quality seems contradictory to this proportion of CM since this is expected to lead to quality and production problems. Segment position 5 is the only segment position with a PM policy in place; this segment is exchanged after 8 weeks in production. However, in 2013 only 37% of the segments on this position reached 8 weeks of production. Moreover, as can be seen from figure 7.1, casting distance does not seem to be a good predictor for the failure probability of a segment, 32% already fails during the first 100 kilometers. Section 2.2 introduced the SCMD. Although it can identify segment failures; it is not suitable for failure prediction. Hence, segment failure predictability is currently non-existent.
figure 7.1: Number of segment exchanges per kilometer interval (July 2012 – July 2014) Reliability+
Figure 7.2 shows that the failure frequency differs per segment position. For example, the average failure frequency of segment position 2 is 166 kilometer, of which 50% of the failure modes are related to breakthroughs (see Appendix VI). Currently, the DSP schedules a planned maintenance interval every 8 weeks, or every 272 kilometers. As can be seen from figure 7.2, the failure frequency is too high to plan segment exchanges in the planned maintenance intervals.
figure 7.2: Average covered kilometers of segment in one production cycle (July 2012 – July 2014)
0" 2" 4" 6" 8" 10" 12" 14" 16" 0"("50" 50"("100"100"("150"150"("200"200"("250"250"("300"300"("350"350"("400"400"("450"450"("500"500"("550"550"("600" Se gm ent'e xc hange s'(no.)' Interval'(km)' 0" 50" 100" 150" 200" 250" 300" Overall" Segment"
posi6on"2" posi6on"3"Segment" posi6on"4"Segment" posi6on"5"Segment" Maintenance"interval"
Costs+
The direct maintenance costs are the costs that are being made from when a segment is transported to GWOC up to the point that the segment is back in the DSP. Over 2013, the direct maintenance cost of segment group 2-5 was €1,33mln. The average direct maintenance cost was around €35.000 per segment but varied significantly (see figure 7.3). These costs consist of two factors: roll related and not roll related costs, which contributed to 41% and 59% respectively.
figure 7.3: Number of segment maintenance projects and direct maintenance costs (over 2013)
The average time required to exchange a segment in the strand is 3,5 hours. However, there is a big deviation (see figure 7.4). The longer intervals (above 10 hours) are related to breakthroughs. The standard cost of an hour of unplanned downtime in the DSP is €12.000. This is based on fixed costs, variable costs, and production loss. However, it must be noted that this is a very modest estimate and that costs might be much higher when upstream and downstream supply chain costs, customer satisfaction, etc. are taken into account. In 2013, this led to a total of 156 hours of production idle time, corresponding to an estimated cost of €1,9mln. As a result, the total costs of segment 2-5 failures was €3.23mln over 2013.
7.3 Identification+of+most+critical+components+
The goal of this step is to identify the components of the MIS (segment group 2-5) that cause the MIS to not reach its functional goal. The functional goal of segment group 2-5 is to support the partly solidified steel slab, to transfer the vertical movement of the slab into the horizontal direction, and to extract heat from the slab to fully solidify the slab.
7.3.1 Identify+failure+mode+with+the+worst+consequences+
The left part of figure 7.5 shows that 59% of the failure modes are related to roll failures. However, breakthroughs are related to the operation of the production line and PM is related to the maintenance policy. Only roll failure, cooling system failure and hydraulic failure are registered intrinsic failure modes. The right part of figure 7.5 only takes these failure modes into account. This shows that roll failures are responsible for 91% of the intrinsic segment failure modes. Therefore, the rolls are identified as a critical component of the segments.
figure 7.5: Percentage of segment exchanges related to failure modes (July 2012 – July 2014)
According to GWOC, the bearings are the limiting factor of roll lifetime. This is confirmed by data analysis: 67% of the roll failures are related to bearing failures. As mentioned above, 91% of the segment failures are related to roll failures. Therefore, a good estimation would be that
91%! ∙ 67% = 61% of the segment failures are related to bearing failures. This can be
explained by the fact that a segment contains 62 bearings (Appendix I) and when one bearing fails, the entire segment fails. For example, from figure 7.1 it was identified that 32% of the segments fail during the first 100 km. When one assumes that 61% of the segments fail because of bearing failures, this leads to a probability of 32%! ∙ 61% = 20% that a segment fails during the first 100 km because of a bearing failure and a corresponding survival probability of 80%. This corresponds to a bearing survival probability of 99,6% during the first 100 km (equation 1). Consequently, a high bearing survival probability can lead to a low segment survival probability.
7.3.2 Identify+performance+of+bearings+ Predictability+and+Reliability+
Despite the criticality of the bearing failures, failure prediction is not possible in the current maintenance system. There are three reasons for this. First, the casting distance is a bad bearing failure predictor; failures are merely the result of random impacts caused by production. Second, rolls are not numbered and not tracked through the closed loop maintenance supply chains. As a result, the influence of different positions in the strand, production events, and the age of bearings are not known. This makes it hard to estimate the failure frequency and MTBF of the bearings. Third, there are no sensors installed in the production process that can monitor bearing condition. The BFID and the SCMD can only be used for bearing failure detection.
Costs+
In 2013, 51 driven rolls and 384 not driven rolls were replaced. This led to a cost of around €0,5mln. Furthermore, 67% of the roll failures were related to bearing failures. Therefore, the direct maintenance cost of failed bearings was €0,5mln! ∙ 67% = €0,34mln. However, this is a modest estimate because bearing failures also drive costs as transport and inspections (see figure 2.4), which are not taken into account. Furthermore, it was identified that 61% of the segment failures are related to bearing failures. Therefore, bearing failures are responsible for
€1,9mln! ∙ 61% = €1,2mln of lost production time. Hence, the sum of direct maintenance costs
and costs of lost production time over 2013 is estimated to be €1,54mln. The performance on the objectives of segments group 2-5 and the bearings is summarized in table 7.2.
table 7.2: Bearing performance on KPIs (over 2013)
KPI Segment 2-5 maintenance Bearing maintenance
Cost €3,23mln €1,54mln
Direct maintenance cost Lost production cost
€1,33mln €1,9mln
€0,34mln €1,2mln
Predictability N/a N/a
7.4 Maintenance+policy+decision+
The next step is to select the combination of maintenance policies that fits the system the best. To do this, the maintenance policy decision tree of Appendix VIII is used. Section 5.2.4 already provided the rationale of these decisions.
Selection+of+best+policies+
Based on the identified objectives and KPIs: cost reduction, reliability and predictability, two maintenance policies have been selected. The first choice is design out maintenance (DOM) since it can continuously improve the segments and the operation of production to improve reliability, which can reduce direct maintenance costs and lost production costs. The second choice is CBM, which can introduce predictability, reduce costs of lost production through planned maintenance, and reduce direct maintenance costs through right on time maintenance. CBM also obtains more data, which can facilitate further improvements.
Use based maintenance (UBM) does not have the potential to introduce predictability because casting distance is a bad predictor for bearing failure because bearing failures are merely the result of random impacts caused by production. Furthermore, the uncertain production environment of the DSP and the changing conditions make that planning maintenance activities on historic data is infeasible.
Gap+between+current+and+best+policy+
From the decision tree stems that reactive maintenance with installed redundancy is the worst choice. Nevertheless, this is the policy that best describes the current maintenance system. This policy is expected to lead to quality problems and loss of expensive processing time, and
