Design automation in industrial order-to-delivery processes : Enabling mass customization of made-to-order products

(1)

Design automation in industrial

order-to-delivery processes

Enabling mass customization of made-to-order products

M˚ans Olander Anton L¨uning

Examiner: Johan Persson Supervisor: Leon Poot

Company supervisor: Albin Mannerfelt

Link¨oping University | Department of Management and Engineering M.Sc. thesis, 30 ECTS | Design and Manufacturing Engineering Spring 2021 | LIU-IEI-TEK-A–21/04039—SE

Link¨oping University SE–581 83 Link¨oping +46 13 28 10 00 , www.liu.se June 9, 2021

(2)

(3)

Copyright

The publishers will keep this document online on the Internet - or its possible replacement - for a period of 25 years starting from the date of publication barring exceptional

circumstances.

The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility.

According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.

For additional information about the Link¨oping University Electronic Press and its

procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.

(4)

Abstract

Today’s manufacturing industry is heading more and more towards a mass customization approach. This enables customers to individually specify product characteristics and unique design features to fulfill their exact needs. A common way to accomplish this is by means of design automation often used together with knowledge-based engineering.

In this thesis, the order-to-delivery process for customized floor gratings at Weland AB is used as a case. This process is currently manual and dependent on several different departments. It results in a time consuming process which is prone to human errors. The purpose is to investigate how design automation can be used and implemented to automate and improve sections of the order-to-delivery process for customized products at an industrial manufacturing company. The objective is to develop a product configurator to automatically generate 3D models and documentation for production and sales support. The configurator succeeded in demonstrating the possible advantages of using design

automation. It showed the possibility to reduce the construction department’s workload and achieve faster time-to-offer for the sales department. Additionally it is concluded that using a product configurator reduces the risk of human errors and opens up possibilities for other improvements, such as reducing material waste in the production.

(5)

Acknowledgements

Firstly, we want to thank our supervisor at XperDi, Albin Mannerfelt, for his guidance and feedback throughout the project. We also want to thank XperDi for giving us the opportunity to write this thesis.

We wish to acknowledge our supervisor Leon Poot and examiner Johan Persson as well as our opponent Frida Lindfors. Thank you for your constructive criticism and support which has elevated the thesis.

All the assistance provided by Weland was greatly appreciated. Thank you for answering all of our questions and providing constructive criticism. A special thank you to Daniel V˚agen¨as for your time and quick replies.

Finally we want to thank both Mehdi Tarkian and Weland for allowing us to use their figures and images in the thesis.

Link¨oping, June 2021 Anton L¨uning and M˚ans Olander

(6)

List of Figures

1 A Pressure Welded floor grating used as a walkway. . . 2

2 Showcasing the flat (rectangular) cross bars for Type A floor gratings. . . 2

3 Example of multiple floor gratings together with a circular recess for a tree. . . . 3

4 L-profile frames used for sides (left) and corners (right). . . 3

5 Current ODP of customized floor gratings at Weland. . . 4

6 Example of how changes to a referenced object can cause errors due to the hierarchical approach in CAD software’s. . . 8

7 Achievable design time allocation using KBE instead of a traditional approach. . 9

8 The four stages of Morphological (left) and Topological (right) transformations. 11 9 Illustration of the project workflow. . . 12

10 Illustration of Scrum workflow. . . 13

11 Illustration of the iteration process for implementing new functions. . . 15

12 Illustration of the structure for the developed product configurator. . . 16

13 Illustration of the user experience flow in the GUI. . . 16

14 Showcasing the flexibility of the floor gratings HLCt model. . . 17

15 Simplified illustration of the inference engine showing the work which the inference engine carries out. . . 18

16 Recess model being instantiated and used to cut a floor grating. . . 19

17 Example of a 3D model for a simplified floor grating. . . 20

18 Example of two floor gratings with frames and support beams. . . 22

19 Example of a 2D drawing for two floor gratings. . . 23

20 Example of a 2D drawing with cutting instructions for two side frame pieces. . . 23

21 Example of assembly instructions for gratings and frames. . . 24

22 Example of assembly instructions for support beams. . . 24

List of Tables

1 Times associated with the product configurator for two different cases. . . 25

(9)

Glossary

API Application Programming Interface CAD Computer-Aided Design

CAE Computer-Aided Engineering CSV Comma-Separated Values DA Design Automation

ERP Enterprise Resource Planning GUI Graphic User Interface

HLCt High Level CAD template

IDE Integrated Development Environment KBE Knowledge-Based Engineering

KBS Knowledge-Based System MC Mass Customization MTO Made-To-Order MTS Made-To-Stock

ODP Order-to-Delivery Process VB Visual Basic

(10)

1 Introduction

Industrial manufacturing companies have throughout history excelled at mass production [1]. This has resulted in cheap and reliable production, in which the products characteristics and design features can be guaranteed. However, customer requests for unique designs to fulfill their individual needs are increasing [2]. Almost perfect is not good enough anymore and companies work hard to adapt to this new, more complex, market.

1.1 Background

A technique which lets customers specify individualized variants of existing products, for a price similar to mass produced ones, is called Mass Customization (MC) [3]. To enable this shift in production technique, manufacturing companies need to adapt to a more flexible production process in which customized products can be produced at nearly the same efficiency as mass production [2]. By switching from the more traditional mass production process of Made-To-Stock (MTS) to Made-To-Order (MTO), companies can offer a wider variety of their products and thereby more easily fulfill customer requirements [1].

The biggest obstacle to overcome is how to implement MC for a similar price as mass production [2]. Customers expect both better and more customizable products without any significant increase in costs. A common solution for achieving this is by incorporating Design Automation (DA) in the process with the potential to not only reduce costs, but also reduce lead times, errors, and manual labor while increasing efficiency [4].

DA enables this through its intent that repetitive, non-creative tasks can be handled by a computer program instead of a human [4]. Using flexible, parametric Computer-Aided Design (CAD) models, or more specifically High Level Cad template (HLCt) models, which are controlled using a separate software together with Knowledge-Based Engineering (KBE) helps accomplish all above [5]. These technical terms are further described in Chapter 2.

1.1.1 Floor gratings

A typical manufacturing industry product, heavily dependant on material processing, is floor gratings. They are metal formed in a grid pattern with the purpose to provide traction while allowing for water drainage and collection of dirt and debris. Floor gratings are popular for use in a wide variety of situations and locations such as in front of building entrances, on walkways, in staircases, and much more [6]. A typical floor grating can be seen in Figure 1.

(11)

Figure 1: A Pressure Welded floor grating used as a walkway. Retrieved from [7]

Floor gratings are divided into two main types, Pressure Welded and Type A [7]. The main difference is in their geometry, or more precisely the cross bars. For the Pressure Welded type, the cross bars are cylindrical and for Type A they are rectangular as seen in Figure 2. This results in some different usage characteristics and are thereby appropriate in different domains.

Figure 2: Showcasing the flat (rectangular) cross bars for Type A floor gratings. Retrieved from [7]

Floor gratings for bigger areas are managed by combining multiple gratings in a specified pattern [8]. This pattern depends on several different factors such as maximum allowed load and weight. Another factor is that in many situations, the gratings require recesses to fit with its surrounding. For example a tree, as seen in Figure 3, or a wall in contact with the gratings area. The recesses can have a lot of different shapes depending on the geometry of the obstacle requiring a recess. To accomplish this, the manufacturer needs to cut the gratings affected and perform some finishing work, for example welding of an edge bar.

(12)

Figure 3: Example of multiple floor gratings together with a circular recess for a tree. Retrieved from [7]

To lay the floor gratings on the desired area, frames and sometimes support beams are required [8]. The frames can have different designs depending on the usage area and location while their dimensions are determined by the corresponding grating [9]. An example of an L-profile frame can be seen in Figure 4.

Figure 4: L-profile frames used for sides (left) and corners (right). Retrieved from [9]

1.2 Companies

This thesis is executed in collaboration with XperDi and Weland. XperDi acts as the product owner and provides the thesis with assistance. Weland is a customer to XperDi and provides the thesis with a case.

1.2.1 XperDi AB

XperDi is a Swedish company located in Link¨oping with expertise in DA. XperDi originates from research in the department of Machine Design at Link¨oping University. Their vision is to minimize repetitive and tedious tasks in manual design processes for industrial companies. They help customers improve their technical innovation by speeding up their development processes. It is mainly done with a modular approach within the field of CAD in which XperDi has developed a generic CAD configurator, named XperDi CAD Configurator (XCC).

(13)

1.2.2 Weland AB

Weland is a Swedish company located in Sm˚alandsstenar with a long history in

manufacturing and sales of floor gratings, spiral staircases, railings, and much more [10]. Pressure Welded gratings is the oldest product they offer and also the backbone to their business. Weland aims to be a flexible company which provides their customers with fast deliveries and design specific requests while ensuring that all quality standards are met. To achieve this, Weland uses a combination of MTS and MTO. More precisely, they work out of big stocks of standardized parts and products which they can then combine and manipulate to produce customized products. This ensures fast deliveries of common items with the ability to handle design specific requests without the need of customized manufacturing machines and methods.

1.3 Thesis case - Customized floor gratings

XperDi frequently helps customers automate and speed up construction work for physical products. They do so by developing task specific configurators which are based on XCC’s core. Therefore, the thesis project is centered around the use of configurators to implement DA in a reliable and robust way.

Weland is interested in improving their Order-to-Delivery Process (ODP) for customized floor gratings with the use of DA. They have therefore tasked this thesis project with developing a product configurator targeting the sales department as direct users.

1.3.1 Current order-to-delivery process

At Weland, the ODP for customized floor gratings is manual and involves several different departments. The process is illustrated in Figure 5. It begins with a customer contacting Weland with a drawing or other instructions for their request. The information is then passed on either directly to the construction department or to production planning,

depending on the requests level of detail and complexity. In some occasions, the order needs to be passed back and forth between construction and production planning. The information passed between employees and departments is usually stored as manually written text and can for more complex orders be supported by simple drawings.

Figure 5: Current ODP of customized floor gratings at Weland.

A quote and confirmation in the form of a text-file, and in some instances also a drawing, is passed back to sales and later to the customer. When the quote and dimensions of the

(14)

product has been verified and accepted by the customer, the order can be sent to production through the production planning department. The process to get the order accepted by the customer can in some occasions require a couple of iterations. When the product in finished, it is sent to the warehouse before being shipped to the customer.

Orders on customized floor gratings can be divided into two separate types: customized individual grating or filling a customized area with gratings. For both of these scenarios, recesses are frequently needed. When filling an area with gratings, the weight per grating and the load each grating can withstand are additional factors to take into account.

1.3.2 Identified areas of improvement

Since many departments are involved in the ODP with information passed between these, there is an issue with miscommunication. This issue also arises from the fact that the information is manually written at each department and passed to the next in the form of text files in an Enterprise Resource Planning (ERP) system. Each time this information is passed on and later interpreted, there is a risk of human error.

The lead times for orders of stocked, non-customized gratings is usually one day. However, when the customer needs customized products, the lead times can vary between a few days to several weeks. This depends heavily on the complexity of the order and the quality of the request from the customer. If the request is of lower quality, it usually results in a need for more communication between departments and the customer. If the order is complex, the need for designers in the construction department is necessary and time consuming. The majority of the time needed for an order is before it is accepted by the customer, resulting in a considerable time-to-offer.

The more complex orders usually involve the need to fill an area with multiple floor gratings. This is currently done mostly by intuition by the experienced designers in the construction department. They base their design decisions, such as grating sizes and support beams placements, from the required maximum grating weight and load capacity. However, since they use intuition and experience they usually add an arbitrary safety factor.

1.4 Purpose and objectives

The purpose is to investigate how DA can be used and implemented to automate and improve sections of the ODP for customized products at an industrial manufacturing company.

The objective is to develop a product configurator, using Weland’s customized floor gratings as a case, with the help of DA. With this configurator, the aim is to:

• Decrease the lead time of the ODP of customized floor gratings. • Prevent communication and product errors.

(15)

1.5 Research questions

Using the product configurator, the following research questions will be answered: 1. How can design automation be used to streamline an industrial order-to-delivery

process for customized products?

2. How can design automation be used to decrease the number of errors for design specific requests?

3. Is a product configurator an effective solution for implementing design automation in an order-to-delivery process for customized products?

1.6 Deliverables

Besides answering the research questions, the thesis aims to deliver a product configurator which can streamline the industrial ODP for customized products. A Graphic User Interface (GUI) will be the primary input method for the salesperson and the output should be 3D models, production documentation and sales support.

For the project owner, the main delivery will be the developed approach for the MC problem and how the ODP can be streamlined and improved. Furthermore, the applications structure and programming code will be of great interest. This also includes specific methods and approaches for solving certain CAD automation problems.

1.7 Delimitations

Delimitations are set to limit the thesis to the given scope.

• The functionality is the main focus of the GUI, meaning design, layout, and user experience are not prioritized.

• Tables for floor grating strength and maximum loads are provided by Weland and integrated into the solution. Therefore, this thesis will not conduct any specific calculations regarding strength of materials.

(16)

2 Theoretical background

DA originates out of several different research areas, all essential for achieving effective DA. The idea of MC can be seen as the the first step in the emergence of DA. Combining this idea with smart methods and tools like parametric CAD modelling and KBE, DA can be feasible and help realize MC.

2.1 Mass customization

The idea of MC can be traced back to the late 80’s and early 90’s [1]. It emerges from mass production but with the intention to deliver products and services suited for each customer’s needs without sacrificing efficiency [1][3]. MC’s key feature is ”the capability to integrate the product varieties derived from the individual customer’s needs with repetition of modularity and the efficiency of mass production, so that the products are affordable due to low product cost achieved by the scale of economy in production” [1].

This results in a shift in design and production standard from MTS to MTO [1]. To support this change, the entire value chain needs to be taken into consideration to satisfy the three essential pillars: time-to-market, variety, and economy of scale.

The focus for MC is on the customers and their individual needs and requirements [3]. These needs and requirements imply new challenges for the production processes which have to be made more flexible and robust. To help accommodate this, modularity and flexibility can be directly designed into the products so that both time and costs for each unique order can be reduced.

2.2 Parametric CAD modelling

3D CAD models are essential in today’s modern engineering and product development [11]. They provide many different advantages but are mainly seen as a way to store design information. Considering it is not uncommon for these models to be quite complex, it is good practice to try and reuse them for different but similar designs. Thus, design reusability is very dependent on CAD reusability.

CAD reusability refers to ”the extent to which CAD data can be effectively altered so it can be used in or adapted to different applications or designs with minimal effort” [11]. Since today’s manufacturing is moving away from mass production and towards design unique user cases, good CAD reusability rapidly grows in importance [12]. Achieving good CAD reusability is however tougher than one might suppose [11]. This is because of how CAD models are built and holds data.

The industry standard for creating these geometric models is feature-based parametric CAD [11]. This entails that the geometry is controlled by non-geometric features called parameters. To store all these features, a record of them is kept in a design tree based on the sequence for which they are created. Hence, all the model’s features are connected hierarchically. Through this, the user has the ability to change design parameters and thereby produce a new final geometry [12].

However, it is in the design tree most of the problems and struggles with redesign and reusability occurs [11]. Since the design tree has a hierarchical approach to connect the

(17)

features with one another, its complexity quickly increases as the number of feature dependencies grow. All these dependencies have to be defined correctly, based on the final geometry, to avoid the entire model being unstable and thereby unable to be reused. Figure 6 shows a common type of error resulting from a change in the design tree which affects the subsequent features.

Figure 6: Example of how changes to a referenced object can cause errors due to the hierarchical approach in CAD software’s.

This robustness problem is handled differently from company to company with methods and guidelines to ensure a certain standardization [11]. However, there is one almost perfect solution to this problem: programming [12]. Programs can precisely identify and select features which are to be modified, provide an exact specification of how the operations are applied, and determine valid combinations of parameter values. This basic approach can be used to ensure stability and robustness for the entire model.

2.2.1 Bottom-up vs top-down modelling

A very common approach for creating CAD models is by developing individual parts separately and then assemble them together [13]. This is usually referred to as bottom-up modelling and is a good solution for creating quick models without complete knowledge of the final shape [13][14]. It is a well suited approach for models where the individual parts and components are standardized and driving the final geometry.

However, these models usually result in a non-dynamic final geometry and even the smallest change to one of the parts might result in an impossible assembly geometry [15]. The

problem is that bottom-up modelling creates a network of references and dependencies not only within each part but also between parts in an assembly. If only one of these references or dependencies changes or is removed, all parts must be manually updated to fit this change.

To cope with this problem, and be able to create dynamic and reusable CAD models, an approach known as top-down modelling is commonly used [13][14][15]. This approach uses a single location, often a so called skeleton model, to control and hold all the important geometry for the entire assembly [15][16]. The individual parts can then be referenced to this model only, resulting in all components adapting accordingly when a change is being made to the main geometry.

(18)

Top-down modelling does however require a better knowledge of the assembly’s final geometry [15]. The final geometry has to be referenced to by all individual parts and components and must therefore be determined at the start of the modelling process.

2.3 Knowledge-based engineering

KBE is commonly referred to as the practice of using software tools to capture engineering knowledge and reusing it effectively. This knowledge is usually captured and stored in a knowledge base where it can easily be accessed throughout the design phase [17]. An inference engine works in symbiosis with the knowledge base and they are together referred to as a Knowledge-Based System (KBS).

KBE is primarily used to automate the design process and reduce non-creative work of the design engineer [18]. This is necessary for a company to provide MC of their products without sacrificing customer requirements or design quality.

Usage of KBE has the objective to streamline the product development by reduction of time and cost involved in the process [18]. This is achieved by automating repetitive tasks in the design phase while simultaneously retaining the design knowledge in order to re-use it, as illustrated in Figure 7. This gives the designer an opportunity to explore a larger area of the design space and hopefully provide a superior product.

Figure 7: Achievable design time allocation using KBE instead of a traditional approach. Adapted from [18]

The most significant advantages from the usage of KBE are listed below [19].

• Reduction of lead time is the main benefit of using KBE. For configurations of products with similar geometric proprieties, different materials or with analytical processes, KBE can be used to reduce time spent on development.

• Product optimization can easily be utilized both in the form of the designer trying different ideas as well as using optimization algorithms to find the best solutions.

(19)

• Knowledge is captured in the product model which is an advantage compared to the knowledge being tied to one individual.

• More time can be used on other parts of the product development process since repetitive and time consuming tasks can be eliminated.

However, using KBE also comes with some disadvantages which are listed below [19]. • Time consuming setup of the product model is the main drawback of using

KBE. For the lead time to be reduced, time needs to be spent early in the product development process to implement KBE.

• Knowledge transfer is a challenge when the system is used as a black box. The user inputs data which is used to generate the output, but the process in between is not understood. To combat this phenomena, it is important to provide an understanding of the whole process for the user.

2.4 Design automation

DA is a broad subject which can refer to any type of system which has the ability to

automate a design task [17]. One key area in DA is geometry-based design automation [17]. It refers to the automatic manipulation of geometric CAD models and is what this thesis mostly will utilize when referring to DA.

To better understand the use of DA in collaboration with CAD, geometrical transformations can be split in two separate groups, morphological transformations and topological

transformations [16].

Morphological transformations refers to the change in shape and form of an object.

According to K. Amadori et al. [16] the level of knowledge kept in an object can be divided into four stages as illustrated on the left in Figure 8. The complexity of the model is tied to each step in the pyramid where the top one is the most complex. On this level, script based relations are used to establish changeable elements which are determined using programming and rules.

Topological transformations, however, are used to instantiate and position items in geometric models [16]. It has three primary uses: adding, removing, or replacing objects or features. Similar to morphological transformations, topological transformations can be divided into four stages, as illustrated on the right in Figure 8. The top step, generic automatic instantiation, uses functions and parameters to determine its instance’s location and

constraints. The instantiation is fully automatic and can both generate and delete geometric instances.

(20)

Figure 8: The four stages of Morphological (left) and Topological (right) transformations. Retrieved from [16]

To achieve efficient and powerful DA, one must implement the top step in both of the pyramids. Combining these top steps, CAD models can be completely flexible and provides the user with the option to generate any type of geometric shape within the products scope [5].

2.4.1 High level CAD templates

An idea to accomplish robust and reusable CAD models, in accordance to the top steps in the morphological and topological transformations pyramids, is by means of HLCt models [16]. Their primary intention is to eliminate non-creative work [16] while enabling design reusability and automation [17].

These flexible CAD models are developed with a set of design and analysis parameters, and then stored in databases [16]. They can be generically instantiated topologically and the geometry can be modified parametrically. By using this approach, high model fidelity can be achieved in the early design phase as well as throughout the entire design process. This enables a top-down approach in which the user has complete control of the final geometry and can easily modify it together with the individual parts and components [16]. This type of modelling is referred to as dynamic top-down modeling [16]. It combines the benefits of both bottom-up and top-down modelling while mostly eliminating their weaknesses.

(21)

3 Methodology

The methodology for this thesis begins with a pre-study, continuing with the development phase of the product configurator, and ending with the project closure. This workflow is illustrated in Figure 9 below. The main part of the project, the development phase, consists of a Scrum inspired iteration process which is more thoroughly explained below in Chapter 3.2 and 3.4.

Figure 9: Illustration of the project workflow.

3.1 Pre-study

The purpose of the pre-study is to build an understanding of the problem and gather necessary data regarding the case. The literature study focuses on academic articles and previous theses, mainly concerning the subjects of DA, KBE and parametric CAD modelling. The findings from the literature study are used as a base for the background, theory and problem formulation but also for the development and validation process of the configurator. More specific data and information is gathered through meetings with the company providing the case, Weland. This knowledge transfer takes place both in official meetings with Weland as well as through discussions with individual employees. Additionally, information about XCC, described in Chapter 3.3.4, and recommended approaches for implementing DA is acquired through discussions with XperDi.

3.2 Scrum

J. Sutherland and K. Schwaber, co-creators of Scrum, define the method Scrum as ”an iterative, incremental framework for projects and product or application development ” [20]. These iterations are the backbone to Scrum and are named Sprints. All sprints begin with a selection of requirements to implement. They end on a pre-defined date, whether the implementation is done or not, with an inspection of the work and a demonstration for the stakeholders. The developed product is to be seen as a completed product, tested and ready to ship, at the end of each sprint. Further sprints improve the product by implementing requirements in order of their importance.

Before the sprints can start, some initial work has to be done to define the problem [20]. An illustration of the Scrum process can be seen in Figure 10. Step one is to define a Product backlog containing all the requested features. The tasks are then divided and dissected into Sprint backlogs. A major theme in Scrum is to inspect and adapt, meaning these sprint

backlogs are adjusted at the beginning of each sprint to adapt it in accordance with the work done in previous sprints.

(22)

Figure 10: Illustration of Scrum workflow.

3.2.1 Implementation

This thesis work implements a downscaled version of Scrum to better suit a project with only two developers. To ensure that the project is heading in the right direction, weekly meetings are held with the product owner and stake holder, XperDi.

Product backlog

The Scrum process starts with a well defined problem statement, or product vision, from XperDi. The problem statement turns into a product backlog using input from the product owner. The product backlog establishes the current problems and how these can be solved with an innovative approach. The resources and assets of the case company, Weland, are also taken into consideration to ensure the final products fidelity.

The product backlog can be seen in Appendix A as a list of product requirements. Sprint backlogs

With the product backlog finalized, the sprints are broken down into sprint backlogs. Each of the sprints have the approach to implement a Minimum Viable Product with just enough to coordinate with XperDi and Weland that their vision of the product will be met.

The project is divided into 6 different sprints with a span between 2-3 weeks depending on their complexity. Each of these sprint backlogs can be seen in Appendix B with a list of tasks and requirements to be completed before the end of the sprint. These backlogs are predetermined but refined at the start of each sprint in accordance to the most important product requirement at the time. They are also adapted in accordance to what was achieved in previous sprints.

(23)

3.3 Tools

This thesis is practically oriented and therefore requires some software tools to succeed. These essential tools are presented below.

3.3.1 Visual Basic .NET

Visual Basic (VB) .NET is an object oriented programming language which runs on the .NET Framework by Microsoft [21]. Thus, VB has access to the .NET class libraries such as Windows Forms GUI library. Windows Forms is a handy tool for creating Windows applications and enables drag-and-drop of visual controls and other similar functionality [22].

3.3.2 Microsoft Visual Studio

Microsoft Visual Studio is an Integrated Development Environment (IDE) which aids the programmer using smart functions such as autofill of variables and debugging [23]. Windows Forms is integrated into Visual Studio and can be controlled using drag-and-drop or programming. The program can handle multiple languages such as VB, C++, Python and Java.

3.3.3 SolidWorks

SolidWorks is a Computer-Aided Engineering (CAE) and CAD software which is primarily used to design and build 3D models [24]. It is a distinguished CAD software in the industry and one of the most popular [25]. SolidWorks, similarly to most other well-known CAD software’s, provides the user with the possibility to control their models parametrically and with user defined rules.

3.3.4 XperDi CAD configurator

XCC is a CAD configurator developed by XperDi with the purpose to ”automate repetitive tasks without limiting the freedom of the design” [26]. The idea is to enable the user to

quickly build complex models and then retrieve essential CAE and ERP data associated with it. This is done by developing HLCt models which XCC then uses and controls in accordance to what the user demands.

3.3.5 API - SolidWorks, XCC and Excel

Both SolidWorks and Excel has support for an Application Programming Interface (API) so that the user can communicate and control the software outside of it [27] [28]. The API can be used to read and write data but also to automate and customize the software.

XCC consists of a GUI, for the users to interact with the application, and a back-end. The back-end can be reached and utilized through an API library. This entails that most of the developed functionality in XCC can be used without its GUI.

3.4 Development workflow

The development workflow of implementing new functions in the product configurator, as seen in Figure 11, is an iterative process. Figure 11 is a detailed view of the iteration process from Figure 9. All the steps included in each iteration can be read about below.

(24)

Figure 11: Illustration of the iteration process for implementing new functions.

3.4.1 Graphical user interface

The GUI is developed in Visual Studio using the programming language VB .NET. To create the visual style of the GUI, Windows Forms library is used. The functionality of the GUI is programmed to execute the desired actions and give the user direct feedback.

3.4.2 Knowledge base

To create the KB, information regarding specific floor gratings from the pre-study is transferred to an Excel document. This acts as a simplified, one-way, ERP with a range of worksheets containing data about different properties for all different floor grating configurations. The second part in developing the KB is the HLCt library modelling.

CAD modelling is carried out in accordance with the theory found in Chapter 2.2 regarding Parametric CAD modelling and 2.4.1 regarding HLCt models. During the construction of the HLCt models, careful consideration is taken in regards to the references. Datum planes, controlled by parameters, are the main references used to ensure robust and flexible CAD models.

3.4.4 Inference engine

The inference engine is the programming code which executes all commands and collaborates with different APIs to gather and generate data. The programming of the inference engine is carried out in Visual Studio using VB.NET. Different classes and subroutines are used to structure the code. Each class has a specific purpose and holds all corresponding variables and functions. The inference engine code for each configurator function is written after the function has been implemented in the GUI and KB.

(25)

4 Results

The result of the thesis is the developed product configurator. This consists of both the approach on how the configurator should be structured as well as the actual developed configurator. How the configurator is structured, and which components it consists of, can be seen in Figure 12. The user solely interacts with the GUI in which information from the KB is communicated and utilized. Based on the input from the user, the desired output documentation is then generated using the inference engine in collaboration with the KB.

Figure 12: Illustration of the structure for the developed product configurator.

The configurator’s different components and how they function can be read about in Chapters 4.1 - 4.3 below. The exported documentation is described in Chapter 4.4.

4.1 Graphical user interface

The GUI guides the user through the configuration of floor gratings. The flow in which this is done can be seen in Figure 13. Screenshots of the GUI’s appearance and different pages can be seen in Appendix C.

Figure 13: Illustration of the user experience flow in the GUI.

The process begins with the user creating a new product order and then choosing to either configure a single grating or to fill an area with multiple gratings. This will open a new page

(26)

in which the user has to determine grating parameters such as grating type, material, size and more. The next step is to add recesses if needed. If the user is configuring an area to fill, additional parameters need to be determined such as max load and max weight of each grating.

All the options for the values to choose from for each parameter is directly communicated and determined from the KB. This restricts the user from inputting parameter combinations outside of what Weland offers. It is done by enabling or disabling certain buttons and

check-boxes, while parameter value options are presented in drop-down lists which are updated in real time.

The grating configuration page has a direct feedback window of the floor grating’s shape, alternatively the shape of the area to be filled, and its dimensions. This page is also used to graphically determine where potential recesses are needed.

When the user is satisfied with the configuration, the export menu can be opened. The different documentation that can be exported can be seen furthest to the right in Figure 12. The options are: 3D models, drawing of the gratings, assembly instructions, frame cutting instructions, and a quote of the order.

4.2 Knowledge base

The KB consists of HLCt models and an Excel document with all the necessary knowledge about the floor gratings and how they can be configured.

4.2.1 Excel

Excel is used as a substitute for an ERP where all grating data is stored. It consists of several worksheets storing all the different versions of floor gratings which Weland offers including all the parameters defining their exact shape and characteristics. There is also a sheet with the maximum allowed load for all sizes of each different floor grating. The product configurator only reads from this document, not automatically updating data or adding order information.

An HLCt library of parametric 3D models was developed to handle all possible cases based of the output from the GUI. The floor grating model is the base model and is always required. Changing the parameters for one of these floor grating models yields different floor gratings using the same flexible HLCt model as seen in Figure 14.

(27)

Each different shape of recess has a separate HLCt model. They contain the volume which is to be removed from the grating, as well as the edge bars which are to be welded onto the grating.

4.3 Inference engine

The inference engine retrieves information from the GUI and KB to automate the creation of necessary production and sales material in the form of 3D models, drawings, and more. Everything, except from the generated quote, is done by automating SolidWorks using its API. An overview of the inference engine can be seen in Figure 15. Each box within the inference engine is further explained below.

Figure 15: Simplified illustration of the inference engine showing the work which the inference engine carries out.

Automation of 3D models

The automation of the 3D models is done in a sequence of automated operations. This sequence and the order of it is described below.

1. Instantiation of the floor gratings HLCt models with the parameters determined by the user in the GUI.

2. Instantiation of HLCt models for recesses, frames, and support beams.

3. Recess cut-outs on floor gratings and support beams. This is done by looping through all recesses and determining all bodies they interfere with and cutting them. If a body is obsolete, it is removed completely.

(28)

Figure 16: Recess model being instantiated and used to cut a floor grating.

The programming code for instantiating the HLCt models with the help of XCC API is written below as pseudocode. The code uses XCC groups to hold and instantiate multiple models together. This is done by looping through all models which are to be instantiated and adding them to the group together with the models parent id. Simultaneously, all parameter values for the models are also modified here.

group = Create New XCC Group() For i = 1 To Models.Count

parentID = Get Model Parent(i) group.Add Model(i, parentID)

For j = 1 To Num Of Parameters(i) group.Modify Parameter Value(i, j) End Loop

End Loop

Instantiate XCC Group(group)

The pseudocode for removing the recesses from the floor gratings can be seen below. The code works by looping through all recesses. For each recess, the volume which is to be removed is used on all gratings. Lastly, the code determines if there are any bodies and gratings left which are obsolete and removes them.

For i = 1 To Recesses.Count

recessVolume = Get Recess Volume(i) For j = 1 To Gratings.Count

gratingModel = Get Grating Model(j)

bodiesToCut = Get Interfering Bodies(recessVolume, gratingModel) Remove Recess Volume(recessVolume, bodiesToCut)

Delete Obsolete Bodies(gratingModel) End Loop

End Loop

Delete Obsolete Gratings()

Automation of drawings and instructions

The drawings and instructions documentation is generated based on simplified 3D models which only represent the outline and thickness of the floor grating. These simplified 3D

(29)

models are created using the same method as the regular 3D models. The only difference is the HLCt models used. An example of this simplified 3D model can be seen in Figure 17.

Figure 17: Example of a 3D model for a simplified floor grating.

The automation of the drawings and instructions documentation all follow the same automated sequence. This sequence and the order of it is described below. 2 and 3 are repeated until there are no more 2D views needed.

1. If needed: Creation of simplified 3D models.

2. Creation of drawing or instruction sheet with a 2D view of the 3D model. 3. Measurements and notes added for relevant features.

The pseudocode for automatically instantiating the models measurements into a drawing can be seen below. Firstly, all of the measurements placed on each side are ordered according to their size, lowest first. Secondly, they are added based on this order. To keep track of the measurements offset, a side counter for each side is used.

Measurements.Sort()

For i = 1 To Measurements.Count

measurementSide = Get Measurement Side(i)

measurementOffset = singleOffset * sideCounter(measurementSide) Add Dimension(i, measurementSide, measurementOffset)

sideCounter(measurementSide)++ End Loop

Automation of order quote

The quote is automatically generated through Microsoft Word via its API. All the necessary data is placed in generic tables. An image of the grating or area to be filled with gratings is exported from the GUI and placed in the quote. Finally, the quote is exported as a PDF file.

4.3.1 Models optimization

Some models are optimized against different suitable algorithms to generate the best solution for each specific case.

(30)

Floor gratings size

The size of each individual floor grating is based on maximum achievable production width and the maximum total weight the customer prefers for each grating. Stronger floor gratings tend to weight more due to thicker steel bars and plates and is therefore generally smaller. The optimization is made by calculating the weight of one square metre of the desired grating. This is then used together with the floor gratings width to determine the maximum length of the grating without violating the maximum allowed weight. Other factors such as the length of the final grating, which has to be greater than a predetermined minimum length, is also taken into consideration in the algorithm.

Support beams position

Support beams are needed everywhere there is a junction between two gratings. They might also be needed to support single gratings. This is determined by the maximum allowed load for the length of a certain type of floor grating. It is done by calculating the max length of the grating which can withstand the load. By comparing this length to the actual length of the floor grating it can be determined how many support beams are needed and where to place them.

Cutting of side frames

The side frames are all being cut out of 2 meter long pieces. To optimize the usage and minimize the material waste of each of these 2 meter pieces, a type of sorting algorithm is used. This algorithm’s purpose is to find the single or added frame lengths closest to 2 meter. It is done by calculating all the different combinations of added lengths which are less than or equal to 2 meter. The frame piece or pieces ensuring the least amount of waste is then removed and the process is repeated for the frame pieces remaining until their is none left. This ensures that the least amount of stocked 2 meter frames are needed.

4.3.2 Product configurator optimization

Several different methods are used to decrease processing time for the configurator. Some tasks and functions, especially when generating files in SolidWorks, are computationally heavy and are therefore optimized to only do what is necessary to create the desired output. Removal of recesses

To cut out recesses from the models in SolidWorks, a function called Indent is used to cut bodies if they interfere. This requires a lot of computation by SolidWorks so instead of doing this for every single body in the model, an algorithm is used to find only the interfering bodies. Thus, the indent function is only performed on the bodies which have been predetermined to interfere.

Export menu

Each order might require different types of output documentation. To speed up this process, an export menu is added to let the user choose the desired outputs and thereby remove unnecessary time for generating documentation not desired.

Data from ERP

When retrieving data from an ERP ,in this case Excel, a slow API call is needed for every single piece of data which is retrieved. A work around for this is to only send out one API call to export the entire database into Comma-Separated Values (CSV) files. These files are optimized to be communicated with in close to real time within the configurator.

(31)

4.4 Finished models and documentation

The models and documents generated from the product configurator are its core output. Examples of these outputs are shown below. The complete drawing and instructions for these examples can be found in Appendix D. A more complex example, showing the potential of the configurator, can be found in Appendix E.

4.4.1 3D models

The fully detailed 3D models of the floor gratings are generated together with corresponding frames and support beams. An example of this model, with a recess on one corner, is shown in Figure 18. All the different steel bars and plates are modeled as individual bodies for maximum configurability and usage.

Figure 18: Example of two floor gratings with frames and support beams.

4.4.2 Drawings

The 2D drawings of the floor gratings are filled with complete measurements of all the dimensions needed for production. An example of this can be seen in Figure 19. These measurements are supported with all the other necessary parameters of the floor gratings written out in the title block.

(32)

Figure 19: Example of a 2D drawing for two floor gratings.

4.4.3 Frame instructions

Instructions on how to cut the side frame pieces are generated as separate sheets on a new 2D drawing. Each sheet has cutting instructions for each individual 2m length of frame needed. The different frame pieces being cut out are numbered to go hand in hand with the assembly instructions below in Chapter 4.4.4. An example of the cutting instructions for one of the 2m side frames can be seen in Figure 20.

Figure 20: Example of a 2D drawing with cutting instructions for two side frame pieces.

4.4.4 Assembly instructions

Assembly instructions are generated as a separate 2D drawing. Each component is numbered in the same way as the regular drawing and frame cutting instructions to show where each component is intended to be placed. An example of this can be seen in Figure 21.

(33)

Figure 21: Example of assembly instructions for gratings and frames.

On a separate sheet, instructions and measurements for laying out the support beams can be found. These measurements are derived from the maximum amount of load each floor grating can handle to ensure structural integrity. An example of this can be seen in Figure 22.

Figure 22: Example of assembly instructions for support beams.

4.4.5 Quote

The quote is generated with all the information needed to confirm the order with a customer. It includes a list of all the parts within the order, prices, a simplified drawing of the

requested order and specifications for the floor grating type. An example quote for a simple order is shown in Appendix D, and one for a more complex order in Appendix E. The quote is generated without the use of SolidWorks.

(34)

4.5 Product configurator efficiency

The speed and efficiency of the product configurator is very dependent on the size and complexity of the floor gratings. The results presented below in Table 1 are for two different orders: one relatively small and simple, one larger and more complex. These two different orders, and the generated production documentation, can be seen in Appendix D (simple order) and Appendix E (complex order) respectively.

Table 1: Times associated with the product configurator for two different cases.

From Table 1, it is clear that the generation of 3D models in SolidWorks is requiring most time. This is even clearer for the bigger and more complex order. The time to generate the drawing and instructions files are increased by 53-482 % ,depending on documentation, from the simple to the more complex order. However, the time to generate the 3D model is increased by 1833 %.

Important to note is that the time to build the customized floor gratings in the GUI is dependent on the user and the communication taking place with the customer. Therefore, the time given above in Table 1 for GUI inputs is approximated effective time the user needs to interact with the GUI. Also, the export times for the production documentation correlates to a computer with mid-range hardware (Intel Core i5 9th Gen and Nvidia RTX 2060) and the configurator being run in debug mode.

(35)

5 Discussion

This chapter is divided in two parts. The first is a discussion of the results from the thesis. The second part is a discussion of the methodology used, what worked in favor of the project and what could have been done differently.

5.1 Results

The developed product configurator has the purpose to both investigate how DA can be utilized in an industrial ODP as well as to provide a concrete example of this. The two most important areas to improve, for the case’s ODP, was to reduce lead-time and prevent errors. Using DA to accomplish this is known to be a common approach. It is also known from MC that it is easiest achieved by improving product flexibility and robustness and thereby satisfy the three essential MC pillars.

The configurator has shown the potential to reduce lead-time and prevent errors by reducing the number of departments involved and information needed to be transferred between these. For the case with Weland’s ODP of customized floor gratings, the configurator can in theory completely remove the need for the construction departments involvement. It also has the potential to reduce and possibly remove the involvement of both the sales and production planning departments. More on this can be read about in Chapter 7.

A large portion of the lead-time today is directly linked to the construction department. Furthermore, most of the errors that might occur in today’s ODP is because of human errors in the transfer of information between the different departments in the process. Consequently, the configurator can be seen as a success in reducing both of these problems with the potential for even further improvements. Being able to exclude the construction department from the ODP is most likely providing the biggest decrease in lead-time. It will also, outside of the ODP, free up time for designers in the department to be spent on more innovative design tasks in accordance to some of KBE’s major advantages.

Another thesis case related challenge is how Weland’s sales department receives and manages floor grating requests from customers. One core function in the product configurator is to help ease this. The process to get a confirmed and approved order from the customer is often time consuming since it might require a couple of iterations and involvement of several different departments. Therefore, the configurator is developed with the intention to reduce the time needed for each iteration.

The sales personnel can directly input the interpreted information from the customer into the GUI and get a direct feedback on all the measurements and other parameters relevant to the order. The configurator can thereafter, in just a few seconds as seen in Table 1, export all this data into a quote which can be sent to the customer immediately. This is a considerable improvement in both time and resources compared to the manual process Weland currently uses.

Weland is also providing their customers with a one day delivery time for their standardized stocked gratings. Optimally implemented MC for the floor gratings should therefore be able to match this delivery time. After an order for customized floor gratings has been accepted by the customer, production documentation needs to be generated. Since the

(36)

be generated within minutes. This is because of how the different core components, in which the production documents are generated, are developed to be completely flexible and robust. For example, the CAD models are developed in accordance to the theory on HLCt models and parametric CAD modelling. Thereby enabling an efficient implementation to achieve maximum flexibility and robustness. These models are then fully controlled and instantiated in accordance to the top steps on the morphological and topological transformations

pyramids in the DA theory.

For the automatic generation of production documentation, it can be seen in Table 1 that the most time-consuming documentation is the 3D models. The time needed is heavily influenced by the number of recesses in the 3D model. The choice of using the indent

function in SolidWorks is based on its robustness compared to other methods and functions. For example, a SolidWorks function called Cavity would only need to be used once on each component making it much more effective. However, Cavity has a history of instability issues resulting in non-robust models. Therefore, it was decided that this thesis would avoid the use of it, sacrificing some time for a more robust product configurator.

Another trivial solution would be to instantiate or generate a sketch and use the SolidWorks function Extruded Cut to create the recesses. This solution would have worked for the specific case of the thesis. However, if one needs to remove a three dimensional shape, for example a tetrahedron, the extruded cut would not be able to carry out that operation. Even though the configurator is developed for a specific case, all the functionality within it is developed to be as generic as possible so that is can be adapted for other cases (products) as well. With this said, the configurator for this specific case can be made faster. However, compared to the manual methods used today, the configurator is significantly faster as the total time can be measured in terms of minutes compared to days or weeks. Furthermore, the time consuming 3D models are not to any use for Weland today. Ignoring the generation of this documentation would decrease the total process time of the configurator significantly. This is even more noticeable for more complex models. The reason for implementing the 3D models is for future production automation at Weland where the 3D models will be of importance.

Apart from the reduced lead-time and errors, the product configurator also adds

functionalities to optimize the output. The ability to have the configurator calculate the optimal solution ensures that the minimum viable solution is found. For example, while designing a solution for an order in which an area is to be filled with multiple gratings, the designers in the construction department uses intuition together with an arbitrary safety factor. This safety factor can in some occasions result in unnecessarily strong designs requiring smaller gratings and additional support beams. Smaller gratings might also result in the need of more cuts in production and thereby more material waste. Instead, the configurator can standardize and optimize these design solutions for floor grating sizes and placement of support beams. It is done by comparing all user determined grating parameters to the max allowed grating weight and load they should carry. This ensures not only less material waste and cost, but also higher reliability in the results because of the standardized method of determining this.

Another area, not used by Weland today, is how the configurator can optimize how the side frames should be cut to minimize material waste. Instead of selling whole 2 meter long side frames for the customers to cut themselves, it is now possible to offer pre-cutting services

(37)

with the help of the configurator. Thus, only the necessary parts need to be shipped to the customer. This adds customer value, saves weight in transport, and makes sure the unused frame pieces can be used for future orders.

A possible area of improvement for the configurator would be to change the way that the HLCt models are instantiated. In the current configurator, the instantiation parameters are determined and set by the inference engine. According to the theory on DA, this could be changed to having the instantiation automatically adapt to its surroundings to reach the top level of the topological transformation pyramid. This would further generalize the results and improve the robustness of the configurator.

There are also possible drawbacks of using a configurator. For example there is a risk of knowledge being lost when certain tasks are automated. Optimally all rules and constraints to make a complete product are integrated into the configurator but there is always a possibility that they have been misunderstood. The advantage for the configurator in this case is that once an error is fixed it should never be a problem again. With that said, software development comes with the risk of software bugs. This is essentially just errors within the program causing it to produce errors or not behave as intended. Identifying these software bugs can be a time consuming process but is needed for a robust and reliable end product.

To set this thesis in an external environmental perspective, it can be argued that the reduction of material waste is a positive improvement. Also, since the risk of errors is decreased, unnecessary production and transportation of products with errors is reduced. Furthermore, the optimizations carried out to find the most effective solutions minimizes the raw materials needed to fulfill each order.

5.2 Methodology

The literature used in the thesis was essential in embedding the given task in an academic setting. It was also important for providing the work with the necessary knowledge to

develop and validate the product configurator. Additionally, the discussions carried out with XperDi regarding XCC and best practices for DA in this thesis context, were essential for speeding up the work and elevating the practical results.

The implementation of Scrum was found useful for the development of the configurator. It kept the focus on important aspects and succeeded in catching miscommunications by the use of minimum viable product. On at least one occasion a function was developed which was found to be unnecessary and Scrum is the reason it was found before too much time was spent on it.

The creation of the product backlog early in the process allowed for a larger focus on developing the functions. When one was finalized it was always clear what the next step would be which made for an effective workflow. Following the product backlog proved to be a successful way of prioritizing which functions to develop and the specific functionalities required.

The utilized tools were found to be effective for the work carried out. XCC in particular greatly aided the controlled instantiation of models into SolidWorks. Without XCC, controlling HLCt models and their parameters while handling different iterations and file

(38)

names would require much more effort and time.

Using Visual Studio to program and structure the code helped to keep track of all the different functions as they were implemented. Windows forms proved to be a useful tool to manage the functionalities and design of the GUI in a quick and easy way. This was important to keep the focus on the DA part of the configurator and perform bare minimum for the GUI to create a working configurator.

The method used to implement new functions in the product configurator worked very well. In some instances, due to the nature of the function being implemented, the implementation order was changed. For example, when implementing certain grating parameters, the data was first added into the KB before being implemented in the GUI. By doing so it was easier to know what exactly needed to be implemented in the GUI for the desired results. However, this approach to implement one function at a time to work fully as desired came with some problems. For example, new functions were always implemented as individual and separate functions and mostly worked by themselves. This was in some cases unnecessary when a similar function already was implemented. The new function could then instead be implemented by reusing the already developed similar function. Preferably by incorporating the new function directly into the old one.

(39)

6 Conclusions

The purpose of this thesis was to investigate how DA can be used and implemented to automate and improve sections of the ODP for a physical product at an industrial manufacturing company. This was achieved through the development of a product

configurator which decreases ODP lead-time, prevents errors, reduces material waste, and adds customer value.

To conclude the thesis, the three research questions will be answered below. 1. How can design automation be used to streamline an industrial

order-to-delivery process for customized products?

In this thesis the focus has been on the sales and construction parts of the ODP for the case. It has been shown that the workload for the construction department could be decreased, with the potential to be removed completely, by the use of DA. Automatic generation of production documentation, which is otherwise done by the construction department, leaves more time for innovative design tasks.

For the sales department, DA through the product configurator gives them the

opportunity to communicate design options and their implications on the order in real time with the customer. Compared to the process of going through the construction department, DA offers a more effective and streamlined process both for the sales personnel and for the customers. This helps reduce the time-to-offer and thereby add value for both the company and customers.

2. How can design automation be used to decrease the number of errors for design specific requests?

It was determined, for the case with Weland’s floor gratings ODP, that most of the errors originates from human interactions (both direct and indirect). The use of DA through the developed product configurator reduces the amount of human interactions necessary in the ODP. DA can therefore, indirectly, be assumed to reduce the number of errors in the ODP for design specific requests. However, this presumes that the method of implementing DA, using a configurator in the thesis case, does not come with other additional errors.

3. Is a product configurator an effective solution for implementing design automation in an order-to-delivery process for customized products?

Based on the results from this thesis work, it is safe to say that a product configurator is an effective solution for the given case with Weland. The configurator reduces lead-time by automating parts of the ODP and reduces material waste both by eliminating human errors and by optimization of the delivered products. Most of the functions implemented in the developed configurator uses a generic approach. Therefore, a product configurator can also be assumed to be an effective solution for implementing DA for general customized product’s ODP.

(40)

7 Further studies

This chapter is divided in two parts. The first, Chapter 7.1, addresses what would need to be done to deliver a complete product. The second part, Chapter 7.2, adds suggestions on further improvements to take the product configurator to the next level, both in terms of DA and more in general how it can elevate the ODP’s effectiveness.

7.1 Finalized product configurator

Some type of verification and validation of the product configurator is of interest before shipping it. One example is to verify the exact time which could be saved for different types of orders using the configurator compared to the current method. The thesis results gives example of time needed to carry out two orders of different complexity. This could be done for real orders at Weland and then compare it to how much time the personnel at Weland needs.

Furthermore, to fully integrate the configurator into the ODP, an implementation of the company’s ERP system is needed. Thereby, the configurator can access information regarding stocked articles, prices, grating configurations and more. Additionally the configurator can output information straight to the ERP to further automate the process. This is an essential part to make the configurator complete which this thesis project has chosen to simplify. As for the HLCt models and generated models, in SolidWorks there is a function called Configurations which is used to manage files which share geometries. Configurations could be integrated to handle different versions of floor gratings as well as different shapes of recesses. This is not essential for the configurators functionality but would reduce the number of CAD-files and significantly improve their ease of maintenance and use.

The thesis project does not take the design of the GUI into consideration. However, for the product configurator to be finalized, some design work would make it easier to use and understand. Using Windows Presentation Foundation instead of Windows Forms might offer more options for the design of the GUI. The GUI would also benefit from usability testing to improve its design and user friendliness.

To improve the usability and structure of the configurator, some sort of order system could be implemented. In some cases, the customer needs to order multiple different floor gratings. Instead of having to run the configurator multiple times, these could all be saved into one single order and exported together.

7.2 Further improvements

Instead of having a customer reach out to sales personnel, the next step would be to let the customer create the order by themselves. A popular industry solution for this is to integrate a product configurator into the companies web page. This could be done using several different approaches. One approach is to build an object handler directly into the GUI and thereby remove the need of an external CAD software. Another approach is to have the web based configurator communicate with an external server and run the CAD software remotely. This approach would probably require less work to implement since all the production documentation could be generated similarly as the developed configurator

Design automation in industrial order-to-delivery processes : Enabling mass customization of made-to-order products