Attention Recurrent Neural Network-Based Severity Estimation Method for Early-Stage Fault Diagnosis in Robot Harness Cable.

Cable is crucial to the control and instrumentation of machines and facilities. Therefore, early diagnosis of cable faults is the most effective approach to prevent system downtime and maximize productivity. We focused on a "soft fault state", which is a transient state that eventually becomes a permanent fault -open-circuit and short-circuit. However, the issue of soft fault diagnosis has not been considered enough in previous research, which could not provide crucial information, such as fault severity, to support maintenance. In this study, we focused on solving soft fault problem by estimating fault severity to diagnose early-stage faults. The proposed diagnosis method comprised a novelty detection and severity estimation network. The novelty detection part is specially designed to deal with varying operating conditions of industrial applications. First, an autoencoder calculates anomaly scores to detect faults using three-phase currents. If a fault is detected, a fault severity estimation network, wherein long short-term memory and attention mechanisms are integrated, estimates the fault severity based on the time-dependent information of the input. Accordingly, no additional equipment, such as voltage sensors and signal generators, is required. The conducted experiments demonstrated that the proposed method successfully distinguishes seven different soft fault degrees.

Current Only-Based Fault Diagnosis Method for Industrial Robot Control Cables.

With the growth of factory automation, deep learning-based methods have become popular diagnostic tools because they can extract features automatically and diagnose faults under various fault conditions. Among these methods, a novelty detection approach is useful if the fault dataset is imbalanced and impossible reproduce perfectly in a laboratory. This study proposes a novelty detection-based soft fault-diagnosis method for control cables using only currents flowing through the cables. The proposed algorithm uses three-phase currents to calculate the sum and ratios of currents, which are used as inputs to the diagnosis network to detect novelties caused by soft faults. Autoencoder architecture is adopted to detect novelties and calculate anomaly scores for the inputs. Applying a moving average filter to anomaly scores, a threshold is defined, by which soft faults can be properly diagnosed under environmental disturbances. The proposed method is evaluated in 11 fault scenarios. The datasets for each scenario are collected when an industrial robot is working. To induce soft fault conditions, the conductor and its insulator in the cable are damaged gradually according to the scenarios. Experiments demonstrate that the proposed method accurately diagnoses soft faults under various operating conditions and degrees of fault severity.

Online and Offline Diagnosis of Motor Power Cables Based on 1D CNN and Periodic Burst Signal Injection.

We introduce a new approach for online and offline soft fault diagnosis in motor power cables, utilizing periodic burst injection and nonintrusive capacitive coupling. We focus on diagnosing soft faults because local cable modifications or soft faults that occur without any indication while the cable is still operational can eventually develop into hard faults; furthermore, advance diagnosis of soft faults is more beneficial than the later diagnosis of hard faults, with respect to preventing catastrophic production stoppages. Both online and offline diagnoses with on-site diagnostic ability are needed because the equipment in the automated lines operates for 24 h per day, except during scheduled maintenance. A 1D CNN model was utilized to learn high-level features. The advantages of the proposed method are that (1) it is suitable for wiring harness cables in automated factories, where the installed cables are extremely short; (2) it can be simply and identically applied for both online and offline diagnoses and to a variety of cable types; and (3) the diagnosis model can be directly established from the raw signal, without manual feature extraction and prior domain knowledge. Experiments conducted with various fault scenarios demonstrate that this method can be applied to practical cable faults.