Injection molding is the most important production method for manufacturing plastic components used in products ranging from cars to medical devices. Although the plastic components themselves are often inexpensive to produce, any defect can lead to expensive errors that can affect the performance or safety of the finished product. Creating a system of early fault detection and continuous process improvement can mean big payoffs for manufacturers.
Real-time release using multivariate data analysis techniques can help reduce error rates for injection molding processes.
This is especially true for the medical device industry, where ultimately patients’ lives and safety rely on the quality and dependability of the products they use. Substantial savings can therefore be made by detecting deviations in time, before they ruin a whole production batch or compromise the quality of the finished product.
Limitations of current quality control methods
For injection molding, the process defect rates may appear to be very low. The processes are stable, and a particular defect may last only two or three cycles out of perhaps hundreds of thousands of cycles a day. The result of a defect may not become apparent until a functional failure in the final product later. To detect a defective component in this situation becomes almost impossible using traditional means of quality control such as visual inspections or measurements. You simply have to find other methods to reach the next level of process optimization.
This is where statistical process control (SPC) comes into to play. For injection molding operations, statistical process control means setting alarm limits on a few key operation variables in order to contain any faulty production steps. Multivariate data analysis (MVDA) based on real-time data monitoring can detect process outliers more efficiently than other methods of SPC by analyzing the correlation structure between multiple dependent variables—often the cause of out-of-spec parts.
Many manufacturers use a system of quality assurance (QA) that relies on univariate analysis (UVA) and monitoring methods. However, in a continuous manufacturing process, these univariate methods can be inadequate because the control charts are reviewed individually for a number of separate processes that happen throughout the operation. This can allow defective parts to be released from one step to the next.
A much more effective way of managing QA for continuous processes is through MVDA.
MVDA is a statistical design tool that allows you to work effectively with large data sets. MVDA considers the correlation structure and relationships that exist between all of the variables being monitored and modeled rather than just the upper and lower limits. While it’s a fairly new approach in the injection molding industry, MVDA is well established in numerous other industries.
Benefits of real-time release
With the huge amount of data that can be collected and analyzed today, companies that want to optimize their operations can implement real-time release for their injection molding processes.
High quality, structured data is a prerequisite for many new technologies that are used to optimize production processes, including multivariate data analysis and real-time release. Though implementing the necessary processes to collect the data can be time consuming, the results are often very rewarding.
Real-time release using MVDA for continuous process manufacturing can provide significant benefits for injection molding operations, such as:
- Improved process consistency
- Increased productivity
- Reduce inventory
- Higher yields and less rework and rejection
- Reduced operator errors
- Lower labor costs
Case study: A more than 90 percent reduction in non-conformances
A global healthcare company decided to optimize their operations by implementing real-time release on their injection molding process. What they did, in short, was first to secure high-quality data. Then, with the use of SIMCA and SIMCA-Q, they created multivariate ideal models of their process. By first building an ideal model, and then running the model in real-time, they were be able to control all the process input data (or factors) that impact the output data, that is, the quality of components.
With real-time release and multivariate data analysis the company:
- noticed a more than 90 percent reduction in the number of non-conformances
- increased throughput by 26 percent, using the same machines and the same production mode, without additional headcount
In the bigger picture, such results can have a fundamental business impact. Especially if you have thousands of machines running 24/7 in plants. By replicating the project successively on more machines all over the world, the company has been able to increase capacity substantially.
Find out more by downloading the case study.