set or not.
Table 2.4 Tool diagnosis example: results of using an RF model.
| Model inputs | Average (%) | Best (%) | Worst (%) |
|---|---|---|---|
| 32 WPT SFs (X/Y axis with level = 4) | 89.5 | 95.0 | 81.2 |
| 32 WPT SFs (X/Y axis with level = 4) + Cutting depth | 90.9 | 96.2 | 85.0 |
| fAE1~fAE4 | 69.1 | 77.5 | 60.0 |
| fAE1~fAE4 + Cutting depth | 81.7 | 90.0 | 76.2 |
2.4.4 Tool Diagnosis using Loading Data
Retaining experienced machine operators is difficult because of poor manufacturing environment. Some forming machines are now equipped with pressure sensors to indicate operators with machine status for compensating their inexperience. To detect failures in a forming process, a pressure sensor (load cell) is attached to a forging die to detect variation in the forging signals of a bolt‐forging machine. Forging failures and loads are generally strongly correlated, but the load distribution may vary with numerous failure modes.
Further, an issue regarding big data exists in identifying failures after long‐term data collection. The cycle time for forming a bolt is only 0.3 s; thus, the signal length is approximately 300 points under the 1 kHz sampling rate of the sensor. The amount of data collected daily is almost 10 MB for one forming machine with data collection performed for 22 hours/day when using eight pressure sensors of four stages (i.e. 8 channels × 1000 data samples/second × 3600 seconds/hour × 22 hours/day). Thus, how to automatically diagnose failure modes from loading data in a forging process becomes a challenge.
Observing Figure 2.32, a forging load (pressure)‐stroke curve demonstrates two intervals: T1 and T2, which can be defined to indicate the characteristics of fastener forming. T1 is an interval representing the time from the die contacting the workpiece to forming after the material has exceeded its yield strength and shown plastic deformation; whereas T2 is the time taken for the cavity to be completely filled after T1.
Figure 2.32 A forging load (pressure)‐stroke curve.
The forging energy during T2 is mainly related to the geometric variation of the die, as illustrated in Figure 2.32. A feature engineering method is used to extract features, including those in time and frequency domains, from the load‐stroke signal in interval T2. For example, avg, std, kurt, RMS, skew, and max are extracted from the time domain and six frequency bands (5 Hz in each bandwidth) are extracted from the frequency domain as defined in Sections 2.3.3.1 and 2.3.3.2.
Additionally, an AEN model is used for reducing the number of forging features; the inputs and outputs are the original and encoded forging load signals, respectively. In this case, the AEN monitors the stability of the bolt‐forming processes and identifies invalid samples, which are mainly affected by the forging pressure.
For example, three failure modes of the bolt‐forming processes and their end products are shown in the upper part of Figure 2.33; these failures usually result from three failure modes including length over‐specification, die notching, and die adhesion. The pressure patterns of the valid process and three failure modes are depicted in the solid and segmented curves in the lower part of Figure 2.33, respectively. Although longer (+0.3 mm) material does not strongly affect the forming process, both die notching and adhesion result in extremely different pressure patterns compared with the original patterns.
Figure 2.33 Failure diagnosis in a forming process.
The single dimension feature in the code of AEN can be used to determine whether invalid bolts have been formed. Because the actual curve of the forging stroke varies, validating the received signal is difficult. Fortunately, the pressure pattern for invalid bolts can be observed using reconstruction of the compressed code by the decoder of AEN with correct results. As shown in the upper part of Figure 2.34, the value for a valid sample is approximately 2.1, whereas that for an invalid sample is approximately −10. The raw data of valid and invalid samples, shown as the dotted curves in the lower part of Figure 2.34, are so similar that the difference could not be identified using a rule or a threshold system.
Figure 2.34 Sample validation using the single dimension feature of the middle layer in the AEN model.
As illustrated in Figure 2.35, an AEN–deep neural network (DNN) is employed to diagnose failures in the forming process. The features extracted by the codes of AEN serve as the DNN model inputs when the reconstructed output X’ of AEN is similar enough to the original input X, which validates that the extracted features of codes are reliable. Consequently, the AEN‐DNN model can not only accurately distinguish valid samples (positive detection rate > 99%) but also correctly diagnose various failure modes (accuracy > 95%). The details of this application case are described in [20] via IEEE DataPort.
Figure 2.35 AEN‐DNN architecture for failure diagnosis.
2.5 Conclusion
This chapter addresses the techniques of data acquisition and preprocessing. For data acquisition, both process data and metrology data have to be collected for developing various intelligent applications. In general, process data consists of sensing signals and manufacturing parameters. As for data preprocessing, the key steps are segmentation, cleaning, and feature extraction. Finally, four practical examples using real‐world data are respectively demonstrated to validate techniques of data acquisition
