Fault classification is vital in smart manufacturing, and convolutional neural network (CNN) has been widely applied in fault classification. But the performance of CNN heavily depends on its learning rate. As the default setting on learning rate cannot guarantee its performance, the learning rate tuning process becomes essential. However, the traditional learning rate tuning methods either cost much time consumption or rely on the experts' experiences, so it is a considerable barrier for the users. To overcome this drawback, this article proposes a CNN with automatic learning rate scheduler (AutoLR-CNN) for fault classification. First, the long short-term memory (LSTM) is used to extract the features of the past loss of CNN. Then, an agent based on deep deterministic policy gradient (DDPG) is trained to automatically control the learning rate for CNN online. Third, the double CNN structure is developed to enhance the stability of the proposed method. The proposed AutoLR-CNN is tested on two famous bearing data sets and a practical bearing data set on wind turbine. The results of AutoLR-CNN are superior to six commonly used baseline learning rate schedulers in Tensorflow. AutoLR-CNN is also compared with other reported machine learning and deep learning methods. The results show that AutoLR-CNN has achieved the state-of-the-art performance in fault classification.

Orchardgrass (Dactylis glomerata L.), is a well-known perennial forage species; however, rust diseases have caused a noticeable reduction in the quality and production of orchardgrass. In this study, genetic diversity was assessed and the marker-trait associations for rust were examined using 18 EST-SSR and 21 SCoT markers in 75 orchardgrass accessions. A high level of genetic diversity was detected in orchardgrass with an average genetic diversity index of 0.369. For the EST-SSR and SCoT markers, 164 and 289 total bands were obtained, of which 148 (90.24%) and 272 (94.12%) were polymorphic, respectively. Results from an AMOVA analysis showed that more genetic variance existed within populations (87.57%) than among populations (12.43%). Using a parameter marker index, the efficiencies of the EST-SSR and SCoT markers were compared to show that SCoTs have higher marker efficiency (8.07) than EST-SSRs (4.82). The results of a UPGMA cluster analysis and a STRUCTURE analysis were both correlated with the geographic distribution of the orchardgrass accessions. Linkage disequilibrium analysis revealed an average r² of 0.1627 across all band pairs, indicating a high extent of linkage disequilibrium in the material. An association analysis between the rust trait and 410 bands from the EST-SSR and SCoT markers using TASSEL software revealed 20 band panels were associated with the rust trait in both 2011 and 2012. The 20 bands obtained from association analysis could be used in breeding programs for lineage selection to prevent great losses of orchardgrass caused by rust, and provide valuable information for further association mapping using this collection of orchardgrass.

With the tendency of decentralization into factories, production scheduling among heterogeneous factories has become a prominent concern in industrial demand response, spurring research on distributed heterogeneous welding shop scheduling problem (DHWSP). Moreover, owing to the inevitable occurrence of uncontrollable system disturbance in practical production environment, the processing time of jobs is uncertain rather than deterministic. Thus, a L-R fuzzy number (LRFN) is introduced to tackle the uncertainty of processing time. Furthermore, in the pursuit of sustainable development, energy efficiency has been a significant emphasis from countries. An effective production scheduling in distributed heterogeneous L-R fuzzy welding shop can optimize both production and energy efficiency, but no related research is reported. Thus, to address this research gap, this paper investigates an energy-efficient distributed heterogeneous L-R fuzzy welding shop scheduling problem (EDHFWSP) with the objectives of minimizing makespan and total energy consumption (TEC). To solve this issue, a self-learning discrete artificial bee colony (SDABC) algorithm is proposed. First, a collaborative initialization is presented to yield excellent initial solutions. Second, a self-learning selection strategy is developed to help solutions select a superior neighborhood structure in employed bee phase. Third, a self-learning variable neighborhood search (SVNS) is designed to adaptively select a neighborhood structure for execution in onlooker bee phase. Fourth, an energysaving strategy is devised to further optimize TEC without affecting makespan. Additionally, to verify the effectiveness of SDABC, extensive experiments are performed to compare SDABC with other 5 optimization algorithms. Experimental results validate that SDABC outperforms its competitors.