Most companies boast of a good quality inspection system. Indeed, quality assurance programs do require many inspections throughout the build cycle. However, inspections merely identify that defects are present, and repair actions on defective hardware involve either a "paper action" (i.e., "used as is," Material Review Board [MRB] disposition, or Request for Waiver [RFW] submission) or physical repair of the defect and resubmission for inspection. Some characteristics of ineffective attempts at preventing defects include (1) management indications that a majority of the problems are "worker caused," (2) a heavy volume of data review to determine causes of poor yields, or (3) strong management reaction to each incident on the factory floor. These approaches virtually eliminate the ability to maintain timely process control, and ensure high incidence of rework/ repair that results in higher production costs and increased process times.
Some companies do strive for prevention of defects by monitoring yields and trends on the production lines. The approach is valid, but is ineffective when there is an overemphasis on the statistics involved with monitoring defects. In many instances no formal defect control system exists, or else it is unregulated or undisciplined. In those situations there will be a lack of concise historical data, poor feedback as to the effectiveness of corrective action taken previously, and a lack of emphasis on the cause of defects.
An extensive volume of data is generated on the factory floor during production. Data typically includes test results, inspection reports, and material discrepancy reports. Analysis of this data and the identification of meaningful trends is essential in identifying the causes creating discrepant hardware. A manual system of data recording and analysis is time consuming and is an inefficient tool in the identification of problem areas on the manufacturing line. Gaps in the data can be anticipated, and the susceptibility of manual analysis to human error can give late and inaccurate yield or defect trends.
An ineffective defect control program can be characterized in several ways. First, there exists a high incidence of rework and scrap. A review of MRB activity will document the level of rework and scrap. A cursory review of the scrap rate may indicate a low level, but a more thorough review of discrepant a hardware, including "use as is" dispositions and the volume of RFW submissions, will give the total picture. A continuous high rate of defect types (i.e., workmanship, part, etc.) without any evidence of decreasing trends is indicative of out-of-control process. Excessive emphasis on the statistical implications of defect or yield rate data, especially in terms of confidence level, is an indication of an ineffective approach to preventive defect control. Such an approach reduces the sensitivity to identifying process problems and underestimates the significance of those problem areas that are identified. A tour of the factory floor is a method to get a quick indication of the adequacy of the defect control program. Lack of visible indications of trends at workstations, and indications that workers are not aware of the types of problems being experienced on the floor and the efforts being undertaken to correct those problems, are symptomatic of difficulty in maintaining preventive defect control.
A management commitment to defect "prevention" is the prime ingredient of a sound defect control program. A management policy on defect control should be evident. The policy should require that management be involved in the review of defect analyses and that the emphasis on defect "prevention" is flowed down to all subcontractors. A management commitment to defect control must be enhanced by a corrective action program committed to defect control. Defect trend information should be obvious on the factory floor. A corrective action team must be established to ensure adequate attention to the causes of defects. Team members should be technical and management personnel familiar with the product being manufactured. The extent of team member's involvement should be triggered by predetermined corrective action thresholds.
For the corrective action team to function efficiently, the volume of inspection data generated on the factory floor must be quickly reduced to identify trends. Such trend information can identify problem areas that are most critical. An automated real time defect reporting/ tracking system is highly recommended to enhance timely identification of problem areas and verification of effectiveness of action taken to correct problems. Such a defect tracking system should correlate defects to (1) location and environment, (2) time in cycle, (3) test step, (4) point in the manufacturing process (as identified on a manufacturing flow diagram), (5) next assembly, (6) symptom description, and (7) similar defects observed in the assembly. In addition, yield and defect thresholds should be set so as to provide meaningful assessment of progress or degradation from the result of corrective action.
There are several good indicators when an effective
defect control program is functioning. First, visible and meaningful
information is posted on the factory floor. There is a distinct sensitivity to
trends, versus waiting for statistical "proof" that a problem exists.
Predetermined corrective action thresholds have been established and action is
being taken based on those thresholds. The corrective action team receives
concise data and is able to identify critical areas that need immediate