Waveforms in Polycarbonate (PC)

Waveforms in Polycarbonate (PC)

This is some text inside of a div block.

In the field of injection molding, where products are continuously formed, the occurrence of defective products is an inevitable challenge. This results in a significant amount of labor being dedicated to manually inspecting each item, a process that is both time-consuming and inefficient.

If it were possible to predict the formation of defects such as silver streaks, sink marks, and foreign material inclusion, it would enable the automatic pickup of defective products predicted to occur in a single shot through collaboration with robots, thus significantly reducing the inspection workload.

MAZIN, Inc. is dedicated to developing AI solutions aimed at solving production challenges on a process-by-process basis. In the injection molding process, we focus on addressing issues such as skill succession and production efficiency improvement. Our goal is to develop algorithms that can be used for defect prediction and automation of molding condition adjustments, among other applications, by conducting various experiments and analyses.

This introduction highlights part of our efforts in experiments and analyses related to such developments.

Overview of the Initiative

In the initiative we are introducing, the objective is to utilize pressure sensors attached to the mold to capture the in-mold pressure during the molding process. By analyzing the time-series waveforms of this pressure data, we aim to identify the occurrence of specific defects such as silver streaks, sink marks, short shots, overpacking/flash, and foreign material inclusion. This approach seeks to determine the feasibility of detecting these specified defects through waveform analysis.


Utilizing polycarbonate (commonly known as PC), which is known for its impact resistance, durability, heat resistance, and moldability, and is used in everything from daily necessities to industrial parts, we conducted injection molding experiments. During these experiments, we generated and analyzed six original feature quantities correlated with the state of the resin flowing inside the mold, aiming to determine various molding defects.

Target Specifications

  • Molded Product: Consumer plastic parts
  • Material: Polycarbonate (commonly known as PC)
  • Measured Variables: Pressure (4 points)


We generated original feature quantities correlated with the state of the resin flowing inside the mold based on the pressure waveforms and conducted analyses.


The results of the initiative allowed us to determine the discernibility of the following molding defects using pressure sensors:

Defect Type / Resin Pressure Sensor

  1. Silver Streaks: △ (Somewhat discernible)
  2. Sink Marks: ◯ (Discernible)
  3. Short Shots: ◯ (Discernible)
  4. Overpacking/Flash: ◯ (Discernible)
  5. Foreign Material Inclusion: ✕ (Not discernible)

Moreover, regarding silver streaks, which are classified as surface defects, the feature quantities developed through this initiative have shown promising signs of being able to distinguish between good products and those with defects.


Conventional wisdom suggests that predicting surface defects from pressure waveforms is challenging. However, it appears possible to predict these defects using feature quantities generated through our unique methods. We will continue to develop our algorithms to achieve higher accuracy and accommodate a wider range of defect types.