Automated surface inspection is an important task in many manufacturing industries and often requires machine learning driven solutions. Supervised approaches, however, can be challenging, since it is often difficult to obtain large amounts of labeled training data. In this work, we instead perform one-class unsupervised learning on fault-free samples by training a deep convolutional neural network to complete images whose center regions are cut out. Since the network is trained exclusively on fault-free data, it completes the image patches with a fault-free version of the missing image region. The pixel-wise reconstruction error within the cut out region is an anomaly image which can be used for anomaly detection. Results on surface images of decorated plastic parts demonstrate that this approach is suitable for detection of visible anomalies and moreover surpasses all other tested methods.

The surface quality of injection molded parts depends on the processing conditions, the cavity surface structure, the runner system, the used polymer and the part geometry. Different replication of the cavity surface structure to the molded part influences its surface function and visual appearance. A descriptive model of the replication process has been developed, taking into account the thickness of the frozen layer during injection. The expected dependencies were proven by systematic injection molding tests with geometric surface micro structures, having depths of 45 or 100 μm and widths from 400 to 2000 μm, that were replicated into high-crystalline polypropylene (HCPP) and polycarbonate (PC). The volumetric structure replication ratio increased with rising mold temperature, melt temperature and cavity pressure. As expected, the mold temperature was dominant for very small micro features. The RMS roughness, determined by confocal microscopy and atomic-force microscopy, was found to be a suitable replication parameter for draw-polished to mirror-finished stochastic cavity surface structures. An abrupt change in wall thickness decelerated the flow front velocity, thus decreased the replicated polymer surface roughness and increased the surface gloss. Moreover, a several micrometer high “surface step” remained, due to the different thickness-dependent shrinkage. This step always ascended from the thicker to the thinner part area. The replication of a mirror-finished mold surface into HCPP was dominated by morphological effects. Local micro shrinkage differences led to micro sink marks, which affected the surface gloss much more than the mold surface structure.

No abstract

The touch-feel sensation of product surfaces arouses growing interest in various industry branches. To entangle the underlying physical and material parameters responsible for a specific touch-feel sensation, a new measurement system has been developed. This system aims to record the prime physical interaction parameters at a time, which is considered a necessary prerequisite for a successful physical description of the haptic sensation. The measurement setup enables one to measure the dynamic coefficient of friction, the macroscopic contact area of smooth and rough surfaces, the angle enclosed between the human finger and the soft-touch surfaces and the vibrations induced in the human finger during relative motion at a time. To validate the measurement stand, a test series has been conducted on two soft-touch surfaces of different roughness. While the individual results agree well with the literature, their combination revealed new insights. Finally, the investigation of the haptics of polymer coatings with the presented measuring system should facilitate the design of surfaces with tailor-made touch-feel properties.

No abstract