Your checklist for the introduction of a geometric component search

The first step has already been taken. You are seriously thinking about systematizing parts usage in your company. What should you think about if you want to introduce such a solution in your company? With the following compilation of important information, we want to give you a few tips and suggestions.

How the processes basically work is already summarized here for you: http://blog.simuform.com/geometrische-aehnlichkeitssuche-was-ist-das

Your 10-point plan for the introduction of geometric component search

In the following article, we provide an overview of important things to consider when implementing a solution for geometric search and geometric similarity search. In doing so, we address the following 10 aspects and have prepared these topics for you:

Point 1: Data formats - CAD formats and more
Point 2: Preparation - classification? No, please.
Point 3: Quality of search and invariance - What does it mean? 
Point 4: Preparation - Automated processes
Point 5: Simple and intuitive use
Point 5.1: Do not forget assemblies - Separate the wheat from the chaff
Point 6: Flexibility in search - How do I sharpen my result?
Point 7: Legacy data - What about the 2D data?
Point 8: Expansion - Is it a one-way street?
Point 9: Licensing - What about occasional use?
Point 10: Commissioning - How quickly can you start?

Point 1: Data formats - CAD format and more

For a geometric search or for a geometric similarity search, a 3D model should be available. But in which format should this be available? 
Your new software for geometric part search should be as independent as possible from the 3D CAD system you use in your company. The CAD data should be able to be processed directly without conversion to an interface format (STEP, STL, etc.). It should also be possible to mix different CAD formats. This makes sense in order to also enable a cross-format search across all parts. This makes it possible, for example, to compare models in SolidWorks format with models in Creo, Catia, or NX format.

Item 2: Preparation - Classification? No, please.

Classifying the models in the dataset is a laborious task. Often, potential customers wrongly assume that a classification of the data is necessary for the introduction of a geometric similarity search according to the above-mentioned principle. This is not correct.

There are also methods in use on the market that require a mandatory classification for the geometric part search. The mandatory classification of the models as preparation of a geometric part search has a number of disadvantages. These include, for example:

• A great effort in creating the class characteristics.
  
  
• Set up a complex set of rules.
  
  
• In most cases, the class characteristics can only be used in design. Sales, AV or purchasing can hardly do anything with them.
  
  
• In many cases, it is necessary that certain design guidelines are followed.

In this context, it is important to implement a system that does not require classification in order to find similar parts in the database.

Point 3: Quality of search and invariance - What does it mean? 

In technical terminology, we speak of a coordinate, symmetry and rotation invariance in search. This means that the 3D drawing parts can be positioned and rotated in space in any way. Their position in space must not have any influence on the search result. This is a very important feature of a powerful system. In practice, it means that the user does not have to worry about how the drawing parts are oriented in space. Finding parts then becomes easier. You can quickly create a simple test yourself here.

Copy any CAD model and rotate it in space in several steps. For each step, save the model. In this way, you create a series of duplicated models that simply take a twisted position in space.

In the second test, repeat this process. You do this then for an axisymmetric drawing part. One of them should be your reference part. Proceed as described above with the other drawing part. In this way, you ensure that a superimposed rotation and symmetry invariance is guaranteed.

Finally, you can repeat this test and move the components in space. Classifying methods certainly do have their limits in this context. Thus you ensure that coordinate, rotation and symmetry invariance is given.

Point 4: Preparation - Automated processes

The success of the introduction of such a solution in the company stands and falls with the acceptance of the users. It is easy to imagine that every process that a user has to take care of is already doomed to death because it is forgotten, because there is no time for it, and because it is simply annoying. Therefore, all processes that are necessary for the permanent use of such a system should be 100% automated in the background. The user does not have to be bothered by this. After all, the software should support him and not be a burden.

Point 5: Simple and intuitive use

Nowadays, no user has the time or the desire to work with complicated software. The handling of the software must therefore be simple and uncomplicated. This is the top priority. No parameters should be set to find similar parts. The result of the search must also be visualized quickly. In any situation, the user must be able to quickly find suitable results:

• In the database.
• With the CAD- PDM- ERP system.
• In combination with external data.
• With digitizing equipment (3D scanner, handheld scanner)

Finding data from the CAD system must be very fast and easy. The CAD system is the most important working tool of the designer. Thus, it should only take the push of a button to find a similar part that the designer has just loaded in his CAD system or is in the process of creating.

Finding data from the inventory must be flexible. It should be possible to find the entry point via all available attributes, master data, and other metadata (order texts from the ERP system, process run cards at the AV, etc.) or documents. It is particularly efficient if the user can find a suitable part without drawing it. It is often assumed that one must first draw a part in order to then search for a similar drawing part. This should not be the case. It should be possible to sift through the data stock, check it, and search for dimensions. For example, you are looking for an L-bracket with 2 holes and a leg length of 45 mm. Thus, the designer should have the flexibility to find suitable components in the stock according to such information without drawing it beforehand.

However, it also happens that a user has received a drawing part from a colleague at another location as STEP, or has downloaded a component from an online standard parts catalog and wants to check ad hoc whether there are similar parts in stock. Even for this use case, one should be able to reach the goal quickly without having to deal with the technology.

Point 5.1: Do not forget assemblies - separate the wheat from the chaff

Do you also want to find a similar assembly? Surely this is a useful thing. Here, methods with classification often reach their limits. Software for geometric similarity search should be able to identify matching similar assemblies as well. Quickly and reliably. Since assemblies are often much larger and more complex than individual parts, care should be taken to ensure that the software used is capable of doing this. Some manufacturers have a low resolution in recognizing similar parts. Then the result of the similarity search for assemblies is not reliable.

You ask yourself, where can this be applied? In practice, it often happens that certain components are manufactured in one piece, while other very similar components are made up of several individual parts. This may be due to production or business reasons. In this case, models of the same or similar drawing parts are available in the inventory as individual parts (cast parts) or as welded assemblies (assembled from several individual parts). The software should be able to identify and visualize the matching hits regardless of the individual part or assembly.

Point 6: Flexibility in search - How do I sharpen my result?

When finding the matching data, every user wants to sharpen his result. Sharpening here means varying the degree of similarity or filtering by selected attributes or component dimensions.

Getting similar parts displayed alone is not always what the experienced user wants. The technical requirements need a sharpening of the search result. For example, the user wants to find similar matches for an L-bracket, which at the same time is made of a certain material and is not wider than 15 mm. At the same time, the degree of similarity should be adjustable in order to be able to think outside the box.

Item 7: Legacy data - What about the 2D data?

What is actually going on with all the 2D legacy data? They may be used less frequently, but the long service life of machinery and equipment means that legacy data inventories need to be accessed. A solution for finding data and information should be able to include 2D data and legacy data. These should be searchable via existing information.

Item 8: Expansion - Is this a one-way street?

The question here is how flexible the system is to include different data sources for the search. Can a PDM system be connected? Is operation with parallel data storage on a file server also possible? What about the connection of an ERP system? Is multilingualism supported? When connecting various data sources, the system should be flexible as well as expandable.

The system can also be expanded with function modules that may be useful for a company in the future. These include process automation, control of the comparison via third-party systems, cross-site use, creation of reports, and independent visualization of data.

The visualization of 3D data is an important topic in companies today. Not everyone has a CAD workstation. But often 3D data must be visualized quickly and superimposed. Thereby it would also make sense to pick up measurements in order to make a qualified decision. Powerful software should offer this. Thus, one can find data in the database, compare them and visualize them easily.

What about the security and confidentiality of the data? Is there a rights system or can the software be synchronized with existing roles and rights systems? Depending on the use case, these points should also be checked.

Item 9: Licensing - What about occasional use?

Software licensing is a horror for every user. Most software products on the market are provided as named user licenses or node-locked licenses. These are workstation-bound licenses, which are ultimately limited in their flexibility. After all, a user does not spend 8 hours a day searching for data and drawing parts. Thus the license is bound to the user but cannot be used effectively in the company. To equip all users with such software becomes then again very expensive. Some use it frequently, others less frequently, etc.

Far more attractive are floating licenses. Occasionally, this type of licensing is also referred to as a network license. These floating licenses can "float" in the network of the entire company. The software can then be used from any workstation that currently has a need for it. Of course, only as many instances of the software can be started at the same time as licenses are available. This form of licensing has a significantly greater range and availability than the workstation-bound licenses mentioned above.

An increase in the floating functionality of the licenses can be achieved by Software Activity Control. In this way, it is checked whether the started instance of the software is also active. So if you have started the software by default when booting the computer, e.g. by the autostart function, but you do not use it actively, then usually a license is wasted. Software Activity Control returns this license to the license manager. If all licenses are out of stock, the user can wait for the next available license. The range and availability of the floating license for Software Activity Control increased by about 20%.

Rent, purchase, or on-premises for short? Each company must decide for itself. A traditional purchase model with annual maintenance is largely antiquated today. With this model, the company usually bears high investment costs for the software and must pay annual maintenance fees. These maintenance fees range from 15% to 20% of the purchase price, depending on the vendor. The company is rigidly bound to the number of licenses and can make changes here unless it buys additional licenses.

The rental model is a small variation of the purchase maintenance model. Here, the purchase price of the software is calculated with the maintenance costs for 48 months, for example, and the total sum is spread over 4 years. Since the rental contract cannot be terminated within the first 48 months, the advantage is only in the accounting. Rental costs can be fully accounted for as operating costs. Investment costs, on the other hand, can be depreciated over 5 years.

The on-premises model, on the other hand, is a very flexible model. It differs from SaaS models in that the software is not run on a cloud solution, but on a hardware/virtual environment of the company, the licensee. The company thus has sovereignty over the data on the system at all times. These are in its own hands and do not leave the company. This is an essential factor, especially in the automotive environment. In the on-premises model, the annual usage price often includes the entire costs for maintenance, updates, upgrades, and remote monitoring. Some manufacturers also include all services that would be required for a major upgrade. In addition, the flexibility of the on-premises model lies in the fact that licenses can also be canceled annually. This means that you are not bound rigidly. This means that a company has fewer concerns with regard to the risk assessment of a new acquisition.

Point 10: Commissioning - How quickly can you start?

It is a decisive factor how much effort you have to put into getting such a solution up and running and keeping it in operation permanently. Does the data preparation lie within your company? Do you have to deal with technological issues or is the complete permanent operation of the overall solution the responsibility of the manufacturer?

What happens when you upgrade your CAD system to a higher version? What impact does this have on your system? A powerful system here is independent of the CAD format and therefore also independent of the upgrade process of the internal CAD systems. A modern solution in the area of geometric search, geometric similarity search or even enterprise search should be robust to such internal changes. It should also be ensured that the connection to your system landscape does not have to be adapted with a software upgrade from the manufacturer. This is often a potential weak point with purely classifying procedures. If the model changes slightly with a CAD upgrade, the entire set of rules may have to be adapted and revised. This could become a permanent construction site.