Migrating to Windchill—Why Reducing Part Count Involves Adding Lots of New Parts

By Lewis Lawrence, Weatherford International

[Editor’s note: In the Spring 2006 issue of Profiles, CAD manager Edwin Muirhead described a project at Weatherford UK to consolidate Pro/ENGINEER part libraries. In this followup article, Weatherford’s global Windchill manager explains about some of the business needs and process changes involved in the overall effort to migrate several business units to a common PDM system.]

Identifying the Problem: What Doesn’t Work?

Weatherford is a large international corporation that provides products and services to the oil and gas industry. The company grew by acquisition and obtained part data in a variety of formats from over 50 different systems. Frequently, the same physical part existed more than once in the dataset. This was particularly true for common parts that have a standard, such as o-rings.

Having to use different part numbers for the same physical object is inefficient. Some of the problem originated from having multiple data management systems. To address this issue, Weatherford has an ongoing project to migrate locations to one common PDM (Windchill) and ERP system. Among the many drivers for this consolidation is smarter inventory management, as well as the economies of scale. Realizing these potential advantages, however, requires not only common systems but also common part numbers.

One function of the part number is to facilitate search and retrieval, and most groups used some sort of smart part numbering scheme to achieve that. But as an enterprise, Weatherford has over one million part numbers and the count grows daily. No intelligent numbering scheme is sustainable with that much data, at least not if it is to be easily understood by all data customers. Similar limitations exist for a description search, as most systems have a limited number of characters available in the description field for a part record.

Theoretical Solution: How Can We Fix This?

If you have ever shopped online, you know that you can find and retrieve an item without knowing its part number. Why can’t finding parts at work be the same? Most retail websites offer a kind of online catalog experience, where items are segregated into a logical hierarchy that you can browse. The term for this kind of data segregation is “classification,” which involves distributing items into classes or categories of the same type.

This is the approach Weatherford uses, defining a classification schema for all its parts and using a next-sequential part number. A classification schema uses a collection of nodes to form a hierarchical structure.

At each node, a collection of attributes forms a template. The templates at each node can differ, so only attributes that suit the parts are applicable.

For efficient data management, consistency is king. For many attributes, it is possible to constrain the list of allowed values to a predefined list, thus ensuring that all data are entered consistently.

In addition, it is possible to apply constraints to attributes on a template so that an attribute value must be populated. This means that if a part is classified on that node, certain attributes have to contain values. These would be “required” attributes. Other attributes on the template might be optional in the sense that if a part is classified on that node, they do not have to contain values and will not be applied to the part. These would be “preferred” attributes.

Defining the Solution: The Theory Seems Sound

Once we defined the classification schema, we needed to slot parts into the appropriate node and to populate their attribute values. This is an iterative process with the classification schema definition. As parts are slotted, the node template evolves to suit the requirements of the parts that use them. Here is an example of one of the parts we saw earlier after it had been classified.

In many cases, it was possible to interrogate the smart part number to obtain the desired attribute values. Where this did not happen, users familiar with the data had to enter the attribute values manually. The combination of classification structure and attribute values provides a convenient and intuitive method for searching. This works in addition to any basic number and description-based searching. Once many like parts are fully classified and their attributes are populated, it is possible to use the attribute values to determine part equivalence.

Implementing the Solution: In the Real World, Things Seldom Go According to Plan

By using the classification data, we identified duplicate parts and passed them to Engineering for “mapping.” Before product structure substitutions could be made, Engineering had to agree that the parts truly were equivalent. This is where the problems started.

Consolidating to one part number means that some groups had to change the number used in “their” product structures. Needless to say, when people are used to relying on “their” part numbers, this is not a popular approach. In addition, the smart part numbers were often used to signify important business information like material specification. Some parts had been in the system for so long that nobody in the organization was exactly certain what the requirements were, and so did not trust the attribute data. Moreover, the part numbers were typically listed in engineering documentation, so updating the product structure required an Engineering Change Notice (ECN).

A typical o-ring, for example, might have eight duplicates that must be consolidated into one. Those eight o-rings might be used in over 10,000 product structures. The costs involved in that volume of engineering work would be hard to justify! In addition to engineering costs, there are the downstream costs related to new part setup in procurement and inventory management of the existing consolidated part numbers.

Fixing the Solution: This Time We Will Do It Better

Changing the part number was clearly not a workable solution, yet somehow we had to get to one single part number. The solution was to add another part identity—a “Part Legacy Number.” This was a new object added to the Windchill model, linked to the part master. Thus, the existing smart part number gave an additional alias identity to the part.

The Part Legacy Number object also holds other data, providing a mechanism to mark a part “Prime” or “Not Prime.” This approach also allows the part to hold multiple legacy part numbers, one primary number, and as many secondary numbers as necessary.

The primary number is the one that always appears on the properties page. The secondary numbers are viewable by mousing over the primary number. In the example shown here, the first six secondary legacy numbers exist as primary numbers on other parts, whereas the last two do not.

The multiple legacy numbers were applied with the following business rules.

• A number can exist as a primary legacy only once.
• A number can exist as a secondary legacy only once.
• A number can exist as both a primary and a secondary, but only on different parts.
• If a part number is added as a secondary, it is immediately marked “Is prime = No.”
• Only certain users with the required permission set can update secondary part numbers.

Having multiple legacy numbers for a part provides the following functionality.

• Multiple smart part numbering schemes can exist without requiring a unique part for each number.
• The Part Legacy Number can be used to search wherever the part number is used.
• The “Is prime = No” provides a controlled, auditable link to the part marked prime.
• The prime part has multiple numbers so that any engineering documentation referring to a legacy number remains correct.
• A part with an “Is prime = No” cannot be added to a product structure. It is automatically replaced with the prime part at the time of edit.
• The legacy numbers can be viewed in a product structure report.

Needless to say, getting this functionality onto a production Windchill system required a lot of customization, testing, and training, but the all effort and costs were justified. The following screenshots show how it works from an end-user perspective.

Deploying the Solution: This Time It Will Work…Honest!

The multiple legacy number functionality provided a way to consolidate part numbers in a cost- effective and sustainable manner. But whose existing part would be marked “prime”? This question had a simple answer—nobody’s!

Instead of trying to work through all the complexities of ownership, specifications, and inconsistently applied standards, it was easier to set up new “enterprise-owned” parts. These were to be new parts for which no engineering group felt ownership. The task of creating the new parts was passed to the procurement team, which worked with a supplier to obtain the required information. The raw data were provided as a spreadsheet and manipulated to generate the required input for creating parts. The following is a sample of how the data looked.

Many of the attribute values were built from the raw data by concatenating values. The material specification, for example, is a Weatherford-specific value that was matched to the appropriate spec supplied by the manufacturer. This information was then used to create a Windchill load file to generate the WTPart objects. A typical load file entry for a part looks like this:

Part;O-ring -001 WE105 (NBR 70 standard) SAE AS568;;A;component;
buy;/Standard Parts/O-Rings;WFord Part in Engineering; Engineering;;RELEASED;ea

CreateOrUpdateIBAValue;Classification;ClassificationNode;
wt.part.WTPart| Weatherford/Standard Parts/Seals/
SAE AS568 (Imperial)

CreateOrUpdateIBAValue;GENERIC DESCRIPTION (NODE SUB-GROUP);O-ring;;;;

CreateOrUpdateIBAValue;COMMON MATERIAL NAME;(NBR, NITRILE);;;;

CreateOrUpdateIBAValue;ELASTOMER HARDNESS (DUROMETER);70 SHORE A;;;;

CreateOrUpdateIBAValue;O-RING MATERIAL SPECIFICATION;WE105 (NBR 70 RECOMMENDED STANDARD);;;;

CreateOrUpdateIBAValue;O-RING STANDARD SIZE DESIGNATION;-001 0.040X0.029 IN (1.02X0.74 MM);;;;

CreateOrUpdateIBAValue;REFERENCE STANDARDS;SAE AS568;;;;

RemoveCSMConstraint;;;;;;;;;

endIBAHolder;;;;;;;;;

A part was created for each size of o-ring in each material specification, meaning that over 8,000 nearly identical parts were created—all with very consistent, easy-to-work-with data. As the parts were loaded, they were assigned sequential part numbers. After load and verification of data, the part data were extracted to an Excel spreadsheet and passed to our CAD administrator so that Pro/ENGINEER models would be created with the correct part numbers. (See Edwin Muirhead’s excellent article for details.)

Once the new parts were available in the system, each engineering group provided a “mapped to” part number for any legacy o-ring part numbers they had. This part mapping provides the basis of a long-term project for part consolidation. The major process steps in this effort are:

• Add secondary number to the prime part and mark legacy parts “not prime.”
• Update all product structures and swap legacy part for prime part.
• Create an ECN to set the state of the legacy part to “Inactive.”
• Release ECN to manufacturing and inventory management so legacy part inventory is transferred to prime part (ensuring open orders, etc. are handled appropriately).
• Set legacy part’s stocking type to “use up” in ERP, ensuring inventory is not created under the legacy part number.

All’s Well that Ends Well: The Proof of a Pudding is in The Eating

While the part mapping and consolidation process is far from perfect, it does work. The technology enables us to maintain smart part numbers without the cost of duplicate parts and users have accepted the system. Now that we have consistent, easy-to-use data, we have developed a culture of re-use rather than create (for o-rings at least). Just as important, Weatherford has preferred supplier agreements with some vendors, and consolidation has resulted in reducing the purchased part costs for o-rings at a number of locations. 

The consolidation process is continually being refined and expanded to other types of parts, with the long-term goal of having a standard parts catalog with only one part number for every part. If this article has got you thinking of a requirement you could meet to improve your organization’s efficiency, just make sure the ends justify the
means.

Lewis Lawrence is the Windchill administrator for Weatherford International, having global responsibility for the company's Windchill deployment. His team is based in Houston, Texas; Aberdeen, Scotland; Shanghai, China; and Edmonton, Canada, supporting over 1500 international users.