[ Home | Links ] Updated 2005-12-27

Design for Assembly

Vincent Chan and Filippo A. Salustri

Introduction

The aim of design for assembly (DFA) is to simplify the product so that the cost of assembly is reduced. However, consequences of applying DFA usually include improved quality and reliability, and a reduction in production equipment and part inventory. These secondary benefits often outweigh the cost reductions in assembly.

DFA recognises the need to analyse both the part design and the whole product for any assembly problems early in the design process. We may define DFA as "a process for improving product design for easy and low-cost assembly, focusing on functionality and on assemblability concurrently."

The practice of DFA as a distinct feature of designing is a relatively recent development, but many companies have been essentially doing DFA for a long time. For example, General Electric published an internal manufacturing producibility handbook in the 1960's as a set of guidelines and manufacturing data for designers to follow. These guidelines embedded many of the principles of DFA without ever actually calling it that or distinguishing it from the rest of the product development process.

It wasn't until the 1970's that papers and books on the topic began to appear. Most important among these were the publications of G. Boothroyd that promoted the use of DFA in industry.

Comparison of Assembly Methods

Figure 1: Relative costs of different assembly methods by type and production volume.


Figure 2: Production ranges for each type of assembly method

Assembly methods can be divided into three major groups.

In manual assembly, parts are transferred to workbenches where workers manually assemble the product or components of a product. Hand tools are generally used to aid the workers. Although this is the most flexible and adaptable of assembly methods, there is usually an upper limit to the production volume, and labour costs (including benefits, cases of workers compensation due to injury, overhead for maintaining a clean, healthy environment, etc.) are higher.

Fixed or hard automation is characterised by custom-built machiner that assembles one and only one specific product. Obviously, this type of machinery requires a large capital investment. As production volume increases, the fraction of the capital investment compared to the total manufacturing cost decreases. Indexing tables, parts feeders, and automatic controls typify this inherently rigid assembly method. Sometimes, this kind of assembly is called "Detroit-type" assembly.

Soft automation or robotic assembly incorporates the use of robotic assembly systems. This can take the form of a single robot, or a multi-station robotic assembly cell with all activities simultaneously controlled and coordinated by a PLC or computer. Although this type of assembly method can also have large capital costs, its flexbility often helps offset the expense across many different products.

Graphically, the cost of different assembly methods can be displayed as in Figure 1. The non-linear cost for robotic assembly reflects the non-linear costs of robots (even small ones cost alot).

The appropriate ranges for each type of assembly method are shown (approximately) in Figure 2.

Assembly methods should be chosen to prevent bottlenecks in the process, as well as lower costs.

Design Guidelines for Manual Assembly

Obviously, the following guidelines depend on the skill of the worker:

• eliminate the need for workers to make decisions or adjustments.
• ensure accessibility and visibility.
• eliminate the need for assembly tools and gauges (i.e. prefer self-locating parts).
• minimise the number of different parts - use "standard" parts.
• minimise the number of parts.
• avoid or minimise part orientation during assembly (i.e. prefer symmetrical parts).
• prefer easily handled parts that do not tangle or nest within one another.

Note that many products do not lend themselves to these guidelines. Many such products are sold as "ready-to-assemble" kits or require that assembly be shifted to cheaper labour markets.

Design Guidelines for Hard Automation

The main different here is that assembly is performed by machines instead of by humans.

• reduce the number of different components by considering
  1. does the part move relative to other parts?
  2. must the part be isolated from other parts (electrical, vibration, etc.)?
  3. must the part be separate to allow assembly (cover plates, etc.)?
• use self-aligning and self-locating features
• avoid screws/bolts
• use the largest and most rigid part as the assembly base and fixture. Assembly should be performed in a layered, bottom-up manner.
• use standard components and materials.
• avoid tangling or nesting parts.
• avoid flexible and fragile parts.
• avoid parts that require orientation.
• use parts that can be fed automatically.
• design parts with a low centre of gravity.

Sometimes it is too difficult to make parts symmetrical, often non-functional features are added to a part to facilitate part feeding, grasping, and orientation.

Design Guidelines for Soft Automation / Robotic Assembly

Compared to humans, robots are extremely inflexible and stupid. However, they can be programmed to do one thing over and over again with high speed and accuracy compared to humans.

• design the part so that it is compatible with the robot's end effector.
• design the part so that it can be fed in the proper orientation.

Evaluation Methods for DFA

It is important to quantify the improvements and goals of DFA. Two methods for DFA quantification considered here are the boothroyd-dewhurst method and the Lucas method.

Boothroyd-Dewhurst Method

This method is based on two principles:

• the application of criteria to each part to determine if it should be separate from all other parts.
• estimation of the handling and assembly costs for each part using the appropriate assembly process.

This method relies on an existing design which is iteratively evaluated and improved. Generally, the process follows these steps:

1. Select an assembly method for each part
2. Analyse the parts for the given assembly methods
3. Refine the design in response to shortcomings identified by the analysis
4. Loop to step 2 until the analysis yields a sufficient design

The analysis is generally performed using some kind of workshee (example shown below)t. Tables and charts are used to estimate the part handling and part insertion time. These "lookup tables" are based on a two-digit code that is in turn based on a part's size, weight, and geometric characteristics.

Non-assembly operations are also included in the worksheet. For example, extra time is allocated for each time the assembly is re-oriented.

Next, parts are evaluated as to whether it is really necessary (in the assembly) by asking three questions:

1. does the part move relative to another part?
2. are the material properties of the part necessary?
3. does the part need to be a separate entity for the sake of assembly?

The list of all parts is then evaluated to obtain the minimum number of theoretically needed parts, denoted by N_m.

* - in column "i", use "1" to represent that a part is essential, and "0" to represent that a part is not essential.

The method then assumes that the assembly time for a part is 3 seconds. With that assumption, the design efficiency can be calculated as:

Design efficiency = (3s x N_m) / T_m.

The charts for this process can be purchased from a company set up by Boothroyd and Dewhurst. As well, they hold workshops and seminars across North America. As this process can be very time-consuming, software is available to help the design engineer. Refer to http://www.dfma.com for further information. A nice screenshoot of the Boothroyd and Dewhurst software is available here.

Lucas Method

The Lucas method is quite detailed, and is described separately.

Basic DFA Guidelines

Here are some basic guidelines for DFA. Generally, you want to start with a concept design and then go through each of these guidelines, decide whether or not it is applicable, and the modify the concept to satisfy the guideline. There is no guarantee that a given guideline will apply to a particular design problem. Many of these guidelines are similar or the same as rules of concurrent engineering.

• Minimise part count by incorporating multiple functions into single parts
• Modularise multiple parts into single subassemblies
• Assemble in open space, not in confined spaces; never bury important components
• Make parts such that it is easy to identify how they should be oriented for insertion
• Prefer self-locating parts
• Standardise to reduce part variety
• Maximise part symmetry
• Design in geometric or weight polar properties if nonsymmetric
• Eliminate tangly parts
• Color code parts that are different but shaped similarly
• Prevent nesting of parts; prefer stacked assemblies
• Provide orienting features on nonsymmetries
• Design the mating features for easy insertion
• Provide alignment features
• Insert new parts into an assembly from above
• Eliminate re-orientation of both parts and assemblies
• Eliminate fasteners
• Place fasteners away from obstructions; design in fastener access
• Deep channels should be sufficiently wide to provide access to fastening tools; eliminate channels if possible
• Provide flats for uniform fastening and fastening ease
• Ensure sufficient space between fasteners and other features for a fastening tool
• Prefer easily handled parts