Optimization of the supply chain to improve the on-time performance

(1)

Optimization of the Supply Chain to improve the On-Time

Performance

Author: A.K.H. Kasteel University of Twente Supervisors: University of Twente

Dr. ir. J.M.J. Schutten Dr. Ir. M.R.K. Mes

PM-Aerotec B.V.

D. Horenberg M. Renting

Master: Industrial Engineering & Management Track: Production & Logistics Management

(2)

(3)

University of Twente

Faculty of Behavioural, Management and Social Sciences

Industrial Engineering and Management

Optimization of the Supply Chain to improve the On-Time Performance

Author:

Dhr. A.K.H. Kasteel

Master:

Industrial Engineering & Management

Supervisors University of Twente:

Dr. ir. J.M.J. Schutten Dr. ir. M.R.K. Mes

Supervisors PM-Aerotec:

Mr. D. Horenberg - Director PM-Aerotec Mr. M. Renting - Project Manager

Date:

October 12, 2020

(4)

(5)

Management Summary

To remain competitive in the aerospace industry, PM-Aerotec needs to maintain a high On-Time Performance.

Due to the growth in revenues in the recent years, PM-Aerotec experiences difficulties to achieve the On-Time Performance level set at 95%. The question therefore arises whether the internal supply chain is capable of handling the growth in revenues. This observed problem of having an On-Time Performance that is consistently too low over the recent years is caused by multiple underlying core problems. The core problems that this research focuses on are the difficulty to estimate the lead time of a new order, the missing insight in the expected production capacity, and the missing forecast on customer demand of new orders specifically. Therefore, the main research question is:

What can PM-Aerotec do to improve the lead time estimation during order acceptance and the expected production capacity estimation to achieve a better On-Time Performance?

First, we analyse the current situation at PM-Aerotec. PM-Aerotec mainly distinguishes between new and recurring orders. There is a significant difference in the internal processes that a new order has to go through in comparison to the recurring order process. There is also a clear difference in the On-Time Performance of new and recurring orders. For new orders, PM-Aerotec has an average On-Time Performance of 61.8% over the last 8 years. For recurring orders, the On-Time Performance over these years is 75.3% on average. The historical performance analysis shows that the On-Time Performance for new orders is fluctuating in the recent years compared to a more stable On-Time Performance for recurring orders. This analysis also illustrates the effect of new orders on the performance of the internal supply chain. We observe a clear relation between the ratio of new orders present in the internal supply chain and the fraction of late orders, as well as a relation between the ratio of new orders and the average lateness of orders.

To improve the supply chain at PM-Aerotec we conduct a literature review on the situation at PM-Aerotec.

With the focus on a Make-To-Order production environment, we create an overview of the available methods to improve the lead time estimation as well as a way to estimate and utilize the expected production capacity to achieve a better On-Time Performance.

Next, we present a conceptual model to solve the core problems at PM-Aerotec. We design this conceptual model to answer the main research question and to solve 2 out of the 3 core problems. The conceptual model excludes forecasting customer demand, as we can predict the lead time of a new order without taking into account future orders. It can however improve the results of this research and is therefore a recommendation for future research.

The conceptual model consists of a simulation model and an artificial neural network. The artificial neural network can predict the lead time of a new order using the order and shop floor characteristics. The neural network predicts the lead time category of the order and we determine 7 lead time categories based on the distribution of lead times of orders at PM-Aerotec: 1-10, 11-15, 16-20, 21-25, 26-30, 31-40, and 41+ working days.

The order characteristics are available in the ERP system of PM-Aerotec whereas the shop floor characteristics are not. With the aim to improve the performance of the neural network, we use a simulation model to generate the missing shop floor characteristics.

We generate two input datasets for the simulation model from the data in the ERP system. One dataset contains the actual orders using the historical data and we generalize the historical data for the other. We refer to these input types for the simulation model as the historical and generalized simulation. In this way, we use the simulation to generate different output and therefore input for the neural network. The historical simulation simulates the exact orders as present in the ERP system at PM-Aerotec to generate the shop floor characteristics for these orders. The generalized simulation generalizes the historical input and uses empirical distributions to determine the order type and production routing upon arrival. Finally, we also use the generalized simulation model for a system performance analysis on the supply chain to determine the robustness of the system’s performance and to identify improvements that can be made according to differences in the performance of the system when altering the nominal system settings. The performance analysis changes the nominal settings meaning the capacity per machine, the processing and setup times, the new orders ratio, the mix of items in the system, the number of arriving orders, and introduces FAI (First Article Inspection) flexibility.

The FAI is a norm in the aerospace industry requiring the approval of a production process on details such as the exact material and machines to produce the item. We illustrate the effect of the FAI flexibility concept per machine group to improve the performance of the supply chain. FAI flexibility allows a product to be produced

(6)

Management Summary ii

on a similar machine, whereas the FAI certification restricts the production of an item to a specific machine.

Currently, PM-Aerotec only certifies one machine for an item for the FAI.

Next to the simulation model, we evaluate 32 neural network configurations varying the input data, weight ini- tialization methods, training algorithms, and activation functions. Next to that, we optimize each configuration by altering the learning rate, batch size and architecture of the neural network. These optimization variables are the hyper-parameters of the neural network. To optimize the hyper-parameters, we use Bayesian optimization which determines the next configuration based on the expected improvement of the performance.

The outcomes of the performance analysis show that an increase in the new orders ratio in the system also mostly decreases the performance of new orders. The reason for this decrease in the performance is the CADCAM department, for which the utilization increases towards 100% for only a slight increase in the new orders ratio.

We therefore recommend to implement capacity planning for the CADCAM department and monitor the new orders ratio in the system to intervene when required. Next to that, we observe that a different mix of items in the system can also improve the performance. If PM-Aerotec shifts its focus to milling 5-axes items and reduces the number of milling 3-axes items, both the On-Time Performance and average lateness can improve.

Finally, introducing FAI flexibility for the milling 5-axes machine group can increase the On-Time Performance, decrease the average lateness, and increase the overall performance of the system.

The evaluation and optimization of the neural network configurations results in the best configuration being the configuration that uses the generalized simulation as input for the order and shop floor characteristics. This configuration is capable of correctly predicting 73.1% of the orders in the train dataset, the dataset that it learns from. Next to that, it correctly predicts 69.0% of the orders in the train dataset that we do not use for the learning process but to validate the performance. Finally, this configuration correctly predicts 63.1% of the most recent orders of PM-Aerotec. These are the orders of the year 2020 that we use as the test dataset.

We continue the analysis on the performance of the neural network for each of the 7 categories of lead times.

For the lead times between 1 and 30 working days, the neural network accurately predicts the orders that actually belong to a category and does not achieve this by over-predicting the orders in one of the categories.

For the orders that have a lead time of 31 working days or more, we observe a lower performance of the neural network. Especially for the 41+ category, the neural network accurately predicts the orders that do belong to this category but also predicts orders from other categories to be part of this category. This shows that the neural network is not capable of recognizing all orders that belong to this category.

Next, we compare the performance of the neural network and the sales department at PM-Aerotec. The neural network correctly predicts 63.1% of the orders whereas the sales department only correctly predicts 56.1%.

Therefore, over all orders in the test dataset the neural network outperforms the sales department. Considering the performance per category, the neural network does outperform the sales department for the lead time categories from 1-30 working days, but for the final 2 categories the sales department performs better.

We also analyze the confidence of the predictions of both the neural network and sales department in terms of the deviations of the predictions. When we allow an error margin of 1 category, the neural network correctly predicts 81.2% of the orders. This shows that about half of the incorrect predictions are within 1 category of the actual category. With this error margin, the sales department correctly predicts 80.2%. This is close to the accuracy of the neural network and therefore illustrates that the neural network is better in precisely predicting the category of an order, but both are capable of closely predicting the lead time category of an order.

To improve the On-Time Performance, we combine the neural network and sales department predictions for the categories in which they outperform the other. When applying this combination on the orders of 2020, we are able to improve the On-Time Performance from 82.5% to 86.1%.

In conclusion, we are able to improve the OTP using the results from the performance analysis and the use of the neural network to predict the lead times of orders of 5 out of the 7 categories. This research is however not able to improve the OTP to the set goal of 95%. Therefore, neither the performance analysis nor the neural network in this research are capable of improving the OTP to the set goal but this research does contribute to eventually achieving it. To continue this improvement towards the 95%, we present the following recommendations for PM-Aerotec. First of all, we recommend the use of capacity planning for the CADCAM department as this is a bottleneck for the new orders. This can also improve the prediction accuracy of new orders by the neural network, as more data is available to learn from. Second, PM-Aerotec should focus on producing milling 5-axes products rather than 3-axes products based on the performance analysis. Third, we recommend PM-Aerotec to implement FAI flexibility for the milling 5-axes machine group. Finally, the possibility to include forecasts on customer demand can improve the prediction accuracy of the neural network and therefore the On-Time Performance.

(7)

Preface

Dear reader,

Hereby I present to you my research to improve the On-Time Performance at PM-Aerotec to obtain the Master’s degree in Industrial Engineering & Management, with a specialization in Production & Logistics Management and an additional specialization in Financial Engineering & Management, at the University of Twente.

This graduation assignment is the final chapter of my study time. During my Bachelors’ degree in Aerospace Engineering at the Delft University of Technology I started to get interested in the business processes in the technical industry. Therefore, I made the decision to start the master in Industrial Engineering & Management to get a grasp of both the technical and business processes. In the end, I am confident that I made the right decision and now I am ready to start my professional career.

I would like to thank everyone at PM-Aerotec for the opportunity to execute my graduation assignment there and for their contribution to this research. A special thanks to my supervisors at PM-Aerotec, Dominic Horenberg and Martin Renting, for their supervision with the focus on challenging me in exploring all the opportunities for this research. Next to that, I would also like to thank Dennis Bakhuis for his support and guidance during the first part of my internship.

Furthermore, I want to thank Marco Schutten and Martijn Mes, my supervisors from the University of Twente.

Marco Schutten for his supervision from the early start until the end of this thesis. His feedback on the analysis of the production environment and the transition towards the solution design consistently pointed me in the right direction to proceed my research in the proper manner. Martijn Mes joined once the area of expertise required to assist with this research became more clear. He provided a lot of feedback on the simulation and neural network given his expertise in these two fields of operations research.

Finally, I would like to my family, friends, and girlfriend for their support throughout my study time.

Arthur Kasteel Hengelo, October 12, 2020

(8)

List of Abbreviations

Abbreviation Definition

OTP On-Time Performance

FAI First Article Inspection CNC Computer Numerical Control

CAD Computer Aided Design

CAM Computer Aided Manufacturing CMM Coordinate Measuring Machine

PDC Plan-de-CAMpagne

KPI Key Performance Indicator

CTP Capable to Promise

MRP Materials Requirements Planning DDAM Due Date Assignment Method

CON Constant

RAN Random

TWK Total Work

SLK Slack

MTO Make to Order

MTS Make to Stock

NOP Number of Operations

CNU Current Number of Jobs

CSU Current Shop Utilization

TWRD Total Work and Random Allowance ANN Artificial Neural Network

SOM Self-Organizing Map

GEP Gene Expression Programming EBP Error Back-propagation

NBN Neuron by Neuron

TPT Throughput Time

ALN Average Lateness

RNS Random Number Stream

SFM Shop Floor Mechanism

FCFS First-Come, First-Serve SPT Shortest Processing Time

LPT Longest Processing Time

EDD Earliest Due Date

JIT Just-in-Time

(9)

1 Introduction

This research is a graduation assignment for the master Industrial Engineering & Management at the University of Twente with the aim to improve the On-Time Performance at PM-Aerotec. PM-Aerotec is part of the PM- Group and produces high precision parts for the aerospace industry. PM-Aerotec has grown a lot over the years and now experiences difficulties to maintain a high On-Time Performance.

This chapter starts with relevant background information on PM-Aerotec in Section 1.1. Next, Section 1.2 elaborates upon the problem as observed by the organization and the identified core problems. After that Section 1.3 presents the research plan. Finally, Section 1.4 discusses the scope of this research.

1.1 Background Information

PM-Aerotec is located in Hengelo (Overijssel) and is part of the PM-Group (Precision Metals). PM-Aerotec produces high precision parts for the aerospace industry. The PM-Group has been founded in 1966 and delivers high precision metal parts to customers all around the world. The PM-Group designs, develops and manufactures high-precision bearings, positioning systems, mechatronical systems, aerospace & military components and modules for both civil and governmental usages.

The high precision parts that PM-Aerotec produces vary heavily due to specific requirements set by their clients.

Next to that, the parts that PM-Aerotec produces also have to meet the international aerospace regulations.

Whilst operating in this very challenging industry, PM-Aerotec aims to achieve a 10% revenue growth annually.

This requires PM-Aerotec to keep the performance of their internal supply chain high to be able to meet both their clients’ and PM’s targets.

Figure 1.1 shows three high precision parts that PM-Aerotec has produced. The parts that PM-Aerotec produces vary from tiny cylinders of just a few millimeters to enormous fuel tanks to be integrated in large-commercial jets. This illustrates the diversity in the parts that PM-Aerotec produces. To be able to produce all these different kinds of parts, PM-Aerotec has a wide-range of production machines. Next to that, each part follows a specific production route to meet the requirements. This all together makes that none of the production processes are the same.

Figure 1.1: Various Parts produced by PM-Aerotec

1.2 Problem Description

PM-Aerotec is experiencing difficulties to maintain a high On-Time Performance (OTP) due to the growth over the recent years. From 2012 to 2019 PM-Aerotec has grown 10.6% yearly on average in terms of revenues and doubled them over the whole period. Figure 1.2 shows the revenues over the years. This illustrates the growth that puts pressure on the internal supply chain.

PM-Aerotec distinguishes between new and recurring orders based on the difference in the processes that these type of orders have to go through in their supply chain. The growth in revenues is also visible in these two categories separately. The revenues from new orders have grown 22.5% yearly on average and therefore almost tripled over the whole period. The revenues from recurring orders have grown 9.6% yearly on average respectively. This increase in revenues for both categories does not only come from the fact that the absolute

(12)

1.2. Problem Description 2

Figure 1.2: Revenues / Year

number of orders has grown (new: 70.0%, recurring: 38.1%) but also by an increase in average revenue per order. The revenues from new orders have grown by 71.5%, whereas for the recurring orders they have grown by 36.5%.

Figure 1.3 shows the revenues from new and recurring orders as the percentage of the total revenues per year.

This ratio has changed slightly from 10% : 90% to 15% : 85% over the years. This ratio shows that the new orders have been and still are only a small part of the revenues generated by PM-Aerotec. This is also reflected in the growth in revenues per category of orders compared to the overall growth of revenues at PM-Aerotec.

Still, the new orders are essential for the continuous growth at PM-Aerotec.

Figure 1.3: % Revenues from New and Recurring Orders / Year

PM-Aerotec questions whether their internal supply chain is able to deal with this increase in revenues. Their clients expect the orders to be delivered on-time whilst meeting the quality requirements. Chapter 2 focuses on the current configuration of the internal processes at PM-Aerotec. The main concern is that the internal lead time of orders is far from optimal due to the coordination between the internal processes.

This concern mainly focuses on the lead time of new orders due to the increase in average revenues of new orders as mentioned before. This average value of new orders also reflects the complexity of the order which will most likely affect the lead time. Still, PM-Aerotec does not only want to optimize the lead time of new orders

(13)

given the observed increase in absolute number of recurring orders. Therefore, PM-Aerotec wants to reduce the average lead time of all orders to be able to meet the industry standards and grow even further as a company.

One of these standards is an On Time Performance of 95%. Figure 1.4 shows the OTP over the recent years for all, new and recurring orders. The average percentage of orders delivered on time over these years is equal to 73.2%, specifically 61.8% for new orders and 75.3% for recurring orders. This shows that for about 25%

of the orders per year the promised due date is not met. Therefore, PM-Aerotec does not meet the standard which they set as a requirement. By reconfiguration of the processes in the internal supply chain an OTP of 95% could be achieved. This would at the same time lead to a decrease in the average lead time of an order at PM-Aerotec.

This observed problem of not being able to meet the OTP requirement is caused by multiple underlying core problems. These problems are identified and Figure 1.5 shows the causal relationships towards the observed problem. Per core problem a brief discussion on the context is given. Chapter 2 provides a more extensive discussion on the observed core problems.

Figure 1.4: On Time Performance of All, New and Recurring Orders

Figure 1.5: Problem Cluster

(14)

• The application process of an additional FAI certificate requires time and capacity

In the case that the machine required for an order based on the First Article Inspection (FAI) certificate is not available in due time, a new FAI application is considered. The request of an additional FAI certificate requires additional lead time for that specific order and potentially other orders as well.

An FAI certificate is a standard in the aerospace industry. A machine that is used to produce a certain order needs to be granted an FAI certification first. Therefore, PM-Aerotec has to produce the part once on the machine they plan to use for the entire order. This first part will be inspected and tested in order for the machine to receive the FAI. After the FAI is granted the entire order gets released to be produced on that machine. Chapter 2 presents additional information about the FAI certification.

In the case of recurring orders it is an option to request an additional FAI if the already approved machine is not available. This however depends on the time until the machine is available again. This request of an additional FAI requires production capacity on the to be granted machine after which the quality engineer has to determine whether the FAI can be granted. Dependent on the time until the required machine with a FAI becomes available, the production planner, CAD-CAM engineers and quality engineer decide which option to go for. In either case, the recurring order gets delayed due to the required FAI certificate in the case the machine is unavailable.

• Large diversity of orders

There is a large diversity in the orders that PM-Aerotec produces for its clients. Each part’s production process is determined according to the requirements set by the customer and aerospace regulations. As mentioned before, PM-Aerotec has produced parts ranging from tiny cylinders to large fuel tanks over the recent years. This large diversification leads to the fact that PM-Aerotec has a very large amount of machines. These machines are capable of producing specific parts, which makes the number of machines that a part can be produced on (next to the FAI requirement) limited.

• Setup time minimisation not used for the production schedule

Currently, the production planner is not minimizing the setup times when creating the production schedule.

This could however reduce the average lead time of orders if orders are grouped that can be produced on a similar machine. This core problem has also been identified and addressed at the headquarters of PM (Antonides, 2019). This core problem is therefore not the focus of this thesis. However, the results of that Master Thesis can still be used at PM-Aerotec as well to improve the production schedule.

• No clear planning or prioritization for the CAD-CAM department

PM-Aerotec has a lot of autonomous machines for which CNC (Computer Numerical Control) production files are required. The CAD-CAM (Computer Aided Design - Computer Aided Manufacturing) department is responsible for this. Chapter 2 presents additional information on the CAD-CAM department and the CNC production files. The CNC files are required for production. Currently, the CAD-CAM department works on a first-come first-serve basis. This implies however that they do not give any prioritization to the orders received from the planner. They only make an exception in the case that production stops due to missing files. In that case, the CAD-CAM department gives priority to this task.

The CAD-CAM department consists of several CAD-CAM engineers who translate the design files from clients into CNC production files. These CNC production files contain the necessary code for the machines to produce the required parts autonomously.

• Difficulty to properly estimate the lead time of a new order

PM-Aerotec’s desire to continue the growth in revenues leads to an issue in order acceptance. This occurs when estimating the lead time of new orders. Currently, the sales manager, production engineer, production planner and CAD-CAM engineers briefly estimate how long it approximately takes to produce the order and use this as an initial input for the lead time to be communicated to the client. The current way of estimating the lead time of a new order is based on the experience with similar orders produced previously. The production engineer determines the routing of the part. Based on the routing, the production planner makes an estimation on the lead time.

If the client demands a somewhat shorter lead time, the sales manager might feel the need to deviate slightly from the estimated lead time to deliver the order. As the estimation is not very accurate this shorter lead time might be feasible. Still, the inability to properly estimate the lead time of a new order could lead to stress further down the supply chain if the actual lead time is longer than expected.

The difference between the expected and realized lead time of a new order also originates from the constant fluctuation between the new vs. recurring orders ratio produced at PM-Aerotec. Due to this changing ratio, the workload in several departments of the supply chain is unbalanced. For instance, the workload of the CAD-CAM department is very high when a lot of new orders have to be scheduled for production.

Overall, a large part of the internal supply chain is put under pressure in the case that the new vs.

recurring orders ratio increases.

(15)

• Customer demand is not forecasted

Currently, PM-Aerotec does not explicitly forecast customer demand for either new or recurring orders.

They assume that the growth in demand on recurring orders continues based on the growth over the recent years. Therefore, they expect a certain number of recurring orders per year. For new orders, PM-Aerotec uses no forecast to predict the customer demand. PM-Aerotec chooses not to as the quotation success rate is a difficult factor to predict. As a consequence, the expected production capacity cannot be properly analysed for a future point in time. The expected production capacity cannot be estimated properly as it does not only depend on the already accepted orders but also on the still to be accepted orders. Next to that, it also realizes fluctuation in the new vs. recurring orders ratio. As mentioned before, the workload in several departments of the supply chain fluctuates due to it. Without forecasting the new orders, the sales manager is not able to estimate the impact of accepting a new order on the workload.

• No clear insight in the expected production capacity

The sales department accepts orders on the short-term to ensure the continuity of growth at PM-Aerotec.

When accepting the orders no clear estimation is made on the expected production capacity. The sales department does consult the production planner for an estimation of the achievable due date. The production planner then judges whether this can be realized given the already scheduled orders. The main issue here is the difficulty to correctly estimate the expected production capacity at a future point in time.

Chapter 2 includes a more detailed explanation of the current processes at PM-Aerotec.

• Execution of production tasks requires specific skills or experience

The production processes at PM-Aerotec are very diverse as mentioned before. The various production tasks require very specific training and/or experience. Due to this, there is a lack in flexibility on how to assign employees to the already planned production tasks. This eventually results in issues with the execution of the created production plan. According to the production manager, the capacity at various production steps is too low to deal with the incoming request of orders.

• No automatic and simultaneous assignment of employees to scheduled production tasks When the planner creates the production schedule, the planner currently limits the number of machines and therefore specifically plans per machine available. However, the planner excludes the planning of employees in the production plan. At a later stage the production manager assigns employees to the scheduled production tasks in the production plan. A consequence of this is that last-minute issues arise when assigning employees to production tasks when insufficient employees are available.

• Rescheduling not possible due to FAI restriction

If production orders are delayed in the production process due to either machine failures or delay in setup of other orders, it is very difficult to find an alternative solution in terms of adjusting the production planning. The production plan is very difficult to adjust as per production order specific machines are required which are granted an FAI for that order, or due to the fact that an additional modelling file would be required. In an attempt to reschedule, The production planner first shifts the scheduled orders towards a later date if acceptable in terms of its due date. Next, the planner considers an additional FAI certification for which the planner consults the CAD-CAM engineers and the quality engineer.

Based on the findings above, part of this research focuses on the new orders. The new orders at PM-Aerotec are partly the cause of disrupting the internal supply chain processes. If there is an imbalance between new and recurring orders, a lot of pressure is put on certain parts of the supply chain.

Over the most recent years, the average percentage new out of all orders varies between 15% and 30%. When this percentage of new orders increases towards the maximum of the observed range, the OTP decreases drastically.

In some cases even below 25%. This shows that an increase in the percentage of new orders is reflected in the OTP of PM-Aerotec. Chapter 2 contains a Historical Performance Analysis which shows the effect on the supply chain of this imbalance of the type of orders .

Next to that, the OTP of new orders is significantly lower than for recurring orders. The FAI requirement for new orders at PM-Aerotec disrupts the internal processes and therefore is part of the research as well. Another topic to focus on is that PM-Aerotec is not able to both estimate the expected lead time of new orders and the expected production capacity properly.

The core problems of this research are:

• The difficulty to properly estimate the lead time of a new order.

• No clear insight in the expected production capacity.

• No forecast is made on customer demand, specifically relevant for new orders.

(16)

1.3. Research Plan 6

The research problem is therefore defined as:

The current estimations of lead times of new orders and the expected production capacity at a future point in time result in an imbalance in the workload in the supply chain over time realizing a non-optimal On Time Performance.

1.3 Research Plan

The observed problem flows from the inadequate estimations of lead times of new orders and the expected production capacity based on the already and the to be requested orders. The current ways of estimating has the consequence that the workload due to the balance between new and recurring orders is fluctuating a lot.

Therefore, a new estimation method for both the lead time of new orders and expected production capacity is designed in order to improve this balance, which will result in a higher On-Time Performance. The improvement of these two estimations is combined as the expected production capacity is related to the lead time of orders.

Next to that, the forecasting of new orders could also improve the estimation of the production capacity. The main research question is defined as follows:

What can PM-Aerotec do to improve the lead time estimation during order acceptance and the expected production capacity estimation to achieve a better On-Time Performance?

To be able to answer the main research question, sub questions are defined:

1. What is the current configuration of the internal supply chain at PM-Aerotec?

1.1. How is the internal supply chain at PM-Aerotec structured?

1.2. What are the current processes in the departments where the identified core problems arise?

1.3. What are the differences in the supply chain flow of new vs. recurring orders?

1.4. What is the historical performance of PM-Aerotec’s supply chain?

Chapter 2 presents the answers to question 1 and the sub-questions. To improve the performance of the internal supply chain one requirement is a clear overview of its structure. Next to that, a detailed understanding of the current processes where the identified core problems arise is another requirement.

For the understanding of the internal supply chain key players are involved to learn from their experience and/or expertise. Another focus of this chapter is on the difference in the supply chain flow between new and recurring orders. A more detailed analysis shows the influence of the new vs. recurring orders ratio on the supply chain performance. Finally, a detailed historical performance analysis on the performance of the supply chain visualizes the identified core problems from historical data available.

2. What is presented in literature regarding the problem at PM-Aerotec and how can the lead time and expected production capacity estimation processes be improved?

2.1. What solution methods are proposed for similar situations as the situation at PM-Aerotec?

2.2. What methods are presented to estimate the lead time of orders?

2.3. What methods are presented to estimate the expected production capacity?

2.4. What methods consisting of a combination of both lead time and production capacity estimation are presented?

Chapter 3 presents the answers to question 2 and the sub-questions. A literature review is conducted to determine the elements of the problem at PM-Aerotec present in literature. Next, the chapter discusses several methods for the estimation of lead times of orders and production capacity, both separately and combined if identified.

3. What solution method from the literature review can solve the situation at PM-Aerotec?

3.1. What elements from literature can be used to model the situation at PM-Aerotec?

3.2. How can the elements identified be integrated into one solution method?

3.3. How to optimize the performance of the solution method and its elements?

(17)

1.4. Scope of Research 7

Chapter 4 formulates the general model to solve the situation at PM-Aerotec. This is based on the findings in Chapter 3. Next, the chapter presents an analysis on the solution methods and the limitations that can be used to solve the problem. Finally, the chapter discusses the optimization of the solution method and its elements.

4. Which solution method is the best for solving the problem at PM-Aerotec?

4.1. Which solution method can estimate the lead time most accurately?

4.2. Which solution method can estimate the production capacity most accurately?

4.3. Which solution method or methods should be used to solve the problem?

Chapter 5 and 6 determine which of the presented solution methods can contribute to solve the problem at PM-Aerotec. The performance of the solution method given the context at PM-Aerotec is analysed and discussed. The performance is analysed using historical data available in the ERP system of PM- Aerotec. These two final chapters answer the main research question. Next, we present a discussion on the implementation of the proposed solution. Finally, we provide an overview of the conclusions and recommendations on the entire research presented in this thesis.

1.4 Scope of Research

The scope of this research in terms of the available data has to be noted. PM-Aerotec has changed their ERP system as of the 1st of January 2020. They have changed the ERP system that they use to monitor their internal supply chain. The new system is called Fujitsu Glovia which replaces Plan-de-CAMpagne (PDC). PDC is a very generic ERP system looking at the possibilities and abilities of ERP systems nowadays. PDC did not view the entire supply chain structure as one but instead treats each department as a separate part with their own working environment. There are barely any links between the actions of the different departments. With Fujitsu Glovia, PM-Aerotec aims to improve the ability to monitor all the ongoing processes in their supply chain by a set of KPIs (Key Performance Indicators) at any arbitrary moment in time.

This research takes this transition into account. As the amount of data that is available in Fujitsu Glovia is limited due to the short period that it is operational, only the (old) data in PDC is used. All data up to the 1st of January 2020 can therefore be used to analyse the supply chain performance. This ensures that a change in the performance due to the transition to Fujitsu Glovia is not taken into account. The data available in Fujitsu Glovia is used to determine the performance of the solution.

All the possibilities in Fujitsu Glovia for PM-Aerotec to monitor and optimize its internal supply chain are currently being implemented. The options available in Fujitsu Glovia are taken into account for the discussion on the eventual implementation of the solution designed in this research.

(18)

2 Context Analysis

This chapter presents and discusses the findings of the context analysis. First, Section 2.1 gives an overview of the supply chain. After the overview the sections consecutively discuss the current processes in the departments where core problems have been identified elaborating on the cause of the problem.

Starting with the order acceptance phase (Section 2.2), the production engineering phase (Section 2.3), production planning (Section 2.4), material requirements planning (Section 2.5), production (2.6), and quality engineering (Section 2.7) are all discussed highlighting the core problems as identified in Chapter 1.

Throughout the discussion of the supply chain elements, each section highlights the difference between the new and recurring orders in terms of their respective supply chain flows. After this discussion Section 2.8 presents a historical performance analysis on and related to the On-Time Performance of the internal supply chain.

Finally, Section 2.9 ends the chapter with the conclusions based on the context analysis.

2.1 Supply Chain Overview

Figure 2.1 gives an overview of the internal supply chain at PM-Aerotec. It shows the flow of an order from the request by a client up to transport of the order after completion. In the case of a new order a parallel flow in the supply chain exist. For new orders, PM-Aerotec has to determine the routing, design the CAD-CAM files and obtain the FAI certification.

When PM-Aerotec receives an order request, they make the distinction between new and recurring orders. For new orders a lead time estimation is required, on which Section 2.2 elaborates. For recurring orders the expected lead time (assuming the same workload) is already known and therefore PM-Aerotec only has to determine when it is able to deliver the requested order. In the case that the client and PM-Aerotec come to an agreement on the due date, the request becomes an actual order. The production engineering department receives the order and starts the preparations for production.

In the case of a new order, the production engineer determines the routing of the order to produce it according to the desired requirements. Based on this the CAD-CAM department designs the CNC production files for the machines in order to produce the requested part. In the mean time the production planner plans the order for production on the machines as determined by the production engineer. After the routing and the CNC files are available, the planner releases the production order and the first completed part is used for the FAI certification.

Section 2.3 elaborates more on the routing, CAD-CAM department, and FAI certification. Section 2.4 discusses the production planning in more detail.

In the case of a recurring order, the production engineer gathers all relevant information and forwards the order to the production planner. The only remaining step before production is the material requirements planning (MRP). Section 2.5 discusses the MRP and Section 2.6 presents an overview of the production processes.

After production the produced parts have to adhere to the quality requirements set by PM-Aerotec’s client and the aerospace regulations. The metrology room checks the dimensions of the part after which the quality engineer gives the ’green light’ on the approval of the produced part. Section 2.7 discusses the quality engineering department which includes the metrology room.

Figure 2.1: Internal Supply Chain Overview of PM-Aerotec

(19)

2.2. Order Acceptance 9

2.2 Order Acceptance

The order acceptance phase is different for new and recurring orders. The process of creating the quote for the client depends on the type of order.

Given a new order request the sales manager, production engineer, material requirements planner, CAD-CAM engineers, and production planner all contribute to the creation of the quote. First, the production engineer determines the routing. Parallelly, the material requirements planner determines the need for new materials or the availability of existing materials. After determining the routing, the CAD-CAM engineers estimate the time required to finish the CNC production files for production. Finally, the production planner estimates the achievable due date for the order request. The sales manager then present the proposed production process with the due date of the order to the client. This process of estimating the lead time of a new order is one of the core problems at PM-Aerotec. PM-Aerotec is currently not able to properly estimate the lead time of a new order due to the following reasons.

First of all, no explicit capacity check is done by both the CAD-CAM department and the production planner.

There is no clear insight in the expected production capacity to determine whether PM-Aerotec is capable to produce the requested order at the required moment in time. Second, PM-Aerotec does not use forecasts to predict customer demand. This restricts the production planner from taking into account possible future orders. Forecasting of new orders could allow PM-Aerotec to improve the estimation of the expected production capacity at a future point in time. Finally, the large diversity and increasing level of complexity in the requested orders over the years makes it very difficult to estimate the lead time correctly, even when the above would be available. This combination of diversity and complexity of the orders realizes a large variance in the lead times of new orders at PM-Aerotec. Section 2.8 illustrates this increase in lead time variance for an increase in the complexity of the order as part of the historical performance analysis.

For a recurring order request, only the production engineer and production planner contribute to the creation of the quote. As the order has been produced before, the routing, materials, and CNC production files from the CAD-CAM engineers are already available. The FAI certification for an order remains valid for 2 years after the last time the order is produced. Therefore, dependent on the time between consecutive orders the FAI certification remains valid for the order. The estimation of the lead time of recurring orders is therefore based on the historical lead time. PM-Aerotec does not thoroughly analyse the difference in workload between the current and the previous time the order is produced.

Fujitsu Glovia: Capable to Promise

The lead time estimation method for new orders as presented above at PM-Aerotec has been used up to the transition to Fujitsu Glovia. As mentioned in Section 1.4, PM-Aerotec has made the transition from PDC to Fujitsu Glovia as of the 1st of January 2020. PM-Aerotec is still in the implementation phase and does not exploit all available modules in Fujitsu Glovia yet. One of these modules is ‘Capable to Promise’ (CTP).

The manual of Fujitsu Glovia (2016) describes the CTP module and its functions. The CTP module is capable to translate a sales order or quote into the earliest possible delivery date based on expected availability of resources and capacity constraints. Inputs for this translation are the BOM (Bill of Materials), routing, and production machines for the to be produced part. The CTP plans a part for the expected due date taking into account constraints such as manufacturing capacity and inventory.

The CTP module has the capability to estimate the lead times of orders. However, the CTP module is currently not able to take into account the specific processes like the CAD-CAM engineering and FAI certification. Next to that, the CTP module bases its estimation on the historical lead times of the order. This data is not available for new orders and therefore the accuracy of the CTP module for the lead time of new orders can be questioned.

Specifically, the CTP module performs two tests to determine the expected due date. These are called the

‘First’ and ‘Alternative’ CTP test. The ‘First’ CTP test uses backward scheduling and requires pegging dates to determine the required MRP starting date of the order to realize the promised due date. Pegging dates are dates set on for instance the ordering of materials to achieve the desired due date. Pegging allows the production planner to use all demand and supply available in the supply chain. The ‘Alternative’ CTP test uses forward scheduling and the current inventory to determine the due date given the current Material Requirements Planning status. The CTP module uses historical information on the lead times of the order but also the ordering of material to determine the lead time of the order, from ordering material to completion of production.

Figure 2.2 shows an overview of the output of the CTP module. It shows the required quantity of the product, current date and the promised due date. It schedules the routing steps in the production schedule as work orders (WO) and creates purchase orders (PO) if the inventory of the required material is insufficient. Finally, it shows an overview of the work orders and purchase orders required to complete the order in time.

(20)

2.3. Production Engineering 10

Figure 2.2: Capable to Promise Example from Fujitsu Glovia (2016)

2.3 Production Engineering

PM-Aerotec requires for every order the routing, the CNC production files, and the FAI certification. First, Section 2.3.1 discusses the routing determination process. Based on the routing, the CAD-CAM engineers create the CNC production files for production. Section 2.3.2 discusses this. In parallel, the quality engineer arranges the FAI certification at the end of the process in order for the client to accept the completed order.

Section 2.3.3 discusses this to conclude the production engineering section.

2.3.1 Routing

The first step in preparing an order for production is determining the routing. The production engineer determines the production route such that the part ends up at the desired requirements as set by the client and aerospace regulations. As there is a large diversity in the orders that PM-Aerotec produces, nearly every order requires a tailored production process and therefore route.

The production engineer determines the routing based on experience. Using the gained knowledge over the years on when for instance products were rejected, the production engineer is capable to determine the required route. Due to the diversity there is no systematic approach for the determination of the route.

For a recurring order the production engineer uses the routing information already available. The production engineer does have to check with the quality engineer whether the FAI certification has expired and therefore has to be renewed. In the case the FAI certification has expired, the production engineer has the option to alter the route given the available machine capacity at that instance.

2.3.2 CAD-CAM Department

The CAD-CAM department provides the CNC production files in order to produce the parts on the autonomous machines at PM-Aerotec. As mentioned before, CAD-CAM is a combination of CAD and CAM. CAD stands for Computer Aided Design and CAM for Computer Aided Manufacturing. The clients of PM-Aerotec provide the CAD files of the part to PM-Aerotec after which the CAD-CAM engineers translate those into CAM files.

The term CAD-CAM is a general term used when automation is introduced in the production processes of companies. CAM files are also often referred to as CNC files due to the structure of the files.

(21)

2.3. Production Engineering 11

Figure 2.3: Illustration of CAD-CAM Software Interface¹

The CAD-CAM engineers at PM-Aerotec create the so-called CNC production files based on the CAD files provided by the clients. CNC stands for Computer Numerical Control which indicates that the CAD files are translated into numbers. These numbers represent the coordinates that the ’cutter’ has to follow when producing the part.

PM-Aerotec produces for instance a fuel tank out of a block of metal. The machine removes material based on the CNC production files to end up at the desired result. Removing material instead of adding it contributes to the strength of the part. This is a requirement set by PM-Aerotec’s clients for a lot of orders. PM-Aerotec therefore provides the option to use the production process called machining. Machining is the term used to describe several material removal processes by means of for instance a cutter.

The CAD-CAM department is becoming more and more important for the new orders at PM-Aerotec. For most new orders, the complexity is of such high level that the use of autonomous machines is beneficial. The workload of this department in the supply chain is constantly fluctuating due to the changing ratio of new versus recurring orders at PM-Aerotec. This has been considered in Chapter 1 as well.

The CAD-CAM department works on a first-come first-serve basis. They create the CNC production files according to the jobs received from the production engineer or production planner. Currently, there is no indication how to properly estimate the time to create a certain CNC file of an arbitrary product. Therefore, it is very hard to estimate the available capacity of this department to handle new orders.

The CAD-CAM department gives no prioritization to the orders except for the case when production stops due to missing CNC files. The CAD-CAM engineers then immediately prioritize this job. Prioritizing the CAD- CAM jobs dependent on for instance the production activities or due dates of parts could reduce the delay of orders. Currently, most delay arises from new orders at PM-Aerotec and this is one of the causes of it.

2.3.3 FAI Certification

The FAI (First Article Inspection) certification is an important process in the supply chain of PM-Aerotec.

Figure 2.4 presents the FAI certification process.

Figure 2.4: FAI Certification Process Flowchart

1https://theedgecutter.com/best-softwares-for-cnc-machine/

(22)

2.4. Production Planning 12

A FAI is a report to verify the design and production of a part manufactured by specific machines using specific material. PM-Aerotec has to provide an FAI certification report to the client which the client will approve upon delivery of the order. The FAI is a requirement for both parties to ensure that the produced part is of sufficient quality and meets the requirements set by industry. The FAI certification is valid for 2 years after the last time a part has been produced.

One of PM-Aerotec’s commitments to clients is that they produce the parts according to the SAE AS9102D requirement set by SAE International (2019). The FAI is part of this requirement set by the aerospace industry.

For the to be granted FAI certificate the production method and the used machines are very important. The production engineer and CAD-CAM engineers have to approve on the combination of routing and CNC files that have been created. Based on that the quality engineer creates a technical drawing of the part.

The quality engineer uses the technical drawing for a pre-FAI before production. PM-Aerotec has introduced the pre-FAI step in their FAI certification process based on rejected series after completion. Previously, the entire FAI certification would be done after completion of production resulting in unexpected flaws in the produced parts. PM-Aerotec now performs the pre-FAI before production to minimize the rejections after production.

The pre-FAI verifies everything in relation to design and the materials used before production. In this way to complete the FAI certification the quality engineer has to ensure that the crucial dimensions are correct after production and that all requirements as set by the client are met.

Using the technical drawing and the pre-FAI procedure, the quality engineer can develop the program for the metrology room parallel to the production of the part. The metrology room has to program the inspection of dimensions after production as this is an automated process.

After the production of one part it is tested to determine whether the entire process is sufficient to obtain the FAI certification. First, the metrology room checks the crucial dimensions. Next, the quality engineer collects all the results of the pre-FAI and FAI certification to determine the overall result. If approved by the quality engineer, the produced parts are complete and are ready for transport. In the case that the quality engineer rejects the produced parts, the production engineer, CAD-CAM engineers and quality engineer discuss the flaws and how to tackle them by either altering the produced parts or by producing completely new parts.

As mentioned before, the FAI certification is valid for 2 years after the production of a part according to the certified process. In the case PM-Aerotec receives a recurring order of which the FAI has expired, they request a delta-FAI for that part. A delta-FAI is a simplified version of the regular FAI certification process. As the machines used are most likely still operational and the materials to be used is the same, the certification can be applied for quite easily again.

The FAI certification requirement for each part restricts the production options at PM-Aerotec. As mentioned in Chapter 1, it limits the production planner in the machines on which orders can be produced due to the FAI requirement. The FAI certification process also takes up time and capacity from multiple departments of the supply chain. Therefore it is debatable in the case of delay of orders whether an additional FAI certification is reasonable given the time and capacity required.

2.4 Production Planning

The production planner schedules the orders for production. Using the routing that the planner receives from the production engineer, the planner schedules the order in the production plan of PM-Aerotec. The planner adds the order to the existing production schedule and uses the planning engine available in the ERP system.

Previously in PDC, the production planner used a forward planning tool with which the planning was renewed once a week by manually emptying and filling the planning again. With the introduction of Factory Planning (part of Glovia), a more advanced planning engine is available to the production planner.

PM-Aerotec uses 3 planning procedures in Factory Planning called Shortforward, Longforward and Longback- ward with which the planning engine creates the production schedule. Shortforward fills up the production schedule for the coming 30 days, taking into account the material resource planning with finite production capacity. The Longforward procedure then schedules for the coming 365 days, allowing materials to be bought in the mean time but still with finite production capacity. The Longforward procedure does not change the already scheduled operations by the Shortforward procedure. Finally, the Longbackward procedure schedules backwards starting at today + 365 days and has the option to change the already scheduled operations. This procedure also has the option to schedule overload and therefore schedules with infinite production capacity.

These procedures use scheduling strategies to prioritize between operations in production. PM-Aerotec has set these strategies for all the procedures the same.

Scheduling strategies in Factory Planning prioritize the operations that are present in the waiting queues.