Optimization of a Thermoforming Machine with Advanced Analytics

Plastic parts are highly used in a variety of products and machines within different industries. Many manufacturers use thermoforming techniques to shape plastics and produce those designed parts.

One of our clients, an international manufacturer of plastic parts for home appliances, uses vacuum thermoforming for its products. The thermoforming machine in operation has the following steps:

  1. Plastic sheet loading
  2. Preheating the sheet in an oven
  3. Heating the sheet in the main oven
  4. Forming the sheet by a negative vacuum pressure produced in the space between
  5. mold and sheet
  6. Cooling the shaped plastic
  7. Removing excess material by a cutter

Challenge

Although the operating thermoforming machine has been retrofitted several times, its technology is old. In addition, the process is highly complex and is affected by various factors such as the raw material quality, sheet temperature during the process, vacuum pressure, oven temperature, temperature, and humidity in the shop floor, just to name a few.

In practice, the highly skilled operators of the machine need to adjust the process control variables. However, due to a high number of variables and dynamic behavior of the process and operation environment, this practice is not optimal and often results in a high rate of production waste and subsequently an increase in cost.

The client wanted to achieve the following goals:

  • Improving Overall Equipment Effectiveness (OEE) by reducing scrap rate
  • Reducing energy consumption

Procedure

In this project, the niologic team had a close collaboration with NEONEX. Together, we managed to connect to PLCs and collect the data from different sensors and actuators. The data was contextualized before transferring to a cloud platform. The historical data of around 600 parameters for four months were gathered for analysis. In the end, more than 1 billion rows of data were collected, cleaned, and stored on the BigQuery data warehouse of Google Cloud Platform (GCP).

We applied correlation analysis and several advanced techniques such as Principal Component Analysis (PCA) to identify critical process parameters. Moreover, our team developed a machine learning model based on a time series neural network algorithm to predict the waste rate of the production, given a set of parameters. Finally, an optimization algorithm was employed to find the optimal parameters of the process for different operational and environmental settings.

Results and Customer Value

The client has now a better understanding of his thermoforming process data and the effect of different parameters on the final product. The engineers and operators are now using the developed model to optimize both the operational parameters and the environmental condition of the shop floor. This has enabled them to reduce the scrap rate by four absolute percentage points, improve OEE and energy consumption. In addition, due to governing parameters identification and complexity reduction, the learning curve of the new operators is improved, and they get trained faster than before.