Читать книгу Electronics in Advanced Research Industries - Alessandro Massaro - Страница 29

Intelligent systems are implemented by AI algorithms, applied also for the intelligent movements of robotic arms [54]. Motors can control robotic arm movements by commands as coded outputs of the AI algorithms. In this scenario, robotic movements are actuated by image processing and image recognition [55]. In this way, image features classification by AI, and IoT data are fundamental to enable automatic processes. A central workstation [56] can manage the automatic processes by receiving control signals and by activating robotic operations. Robotic manipulation is an important topic in industry automation involving application fields such as self‐adaptive robotic fingers optimized for collaborative robots [57]. The AI engine can optimize the production parameters. The parameter configuration can be performed off‐line [58] or in‐line. The in‐line parameter setting by AI, is an issue for the modern adaptive solutions: auto‐calibration in real time of machine working parameters can be catastrophic if the AI model fails. For this purpose, it is fundamental to construct robust AI models. A more complex industrial environment involves multiple workers, worker tracking, and security aspects [59]. In this scenario the auto‐adaptive solution is applied also to check dangerous positions of workers by stopping machines in the case of alerting conditions, by applying image vision techniques. Image vision is implemented also for robot guidance solutions [11]. Production process simulations improve robotic production. Simulations are performed in static (time independent) or dynamic (time dependent) conditions [60], and are useful to define adaptive conditions. Adaptive solutions are oriented on reconfigurable applications supporting complex positioning tools [61], and robotic capabilities including sensing, production intelligence, and motion. Robots automatically follow instructions by a standard program or generate in real time mechatronic actions by processing sensor data. To optimize the formulation of instructions, image vision techniques are potentially applied together with AR tools, by improving the monitoring of the correct product assembly, by detecting and predicting defects detection, and by adapting actuations. The input data such as historical production data, data sensors and digitized information, are processed for a dynamic parameter setting and for the formulation of new high‐performance programs. Historical production data are typically collected into a big data system able to contain massive data. Digitized information is coded into a dataset to be processed by the AI engine. The PLC is typically interfaced to an electronic board reading a standard program (standard parameter set), or actuating command coming from a protocol, by means the decoding of the AI outputs. The production process control is usually performed by image vision techniques and by IoT sensors placed inside the machines (internal IoT) or outside (external IoT). The architecture describing all the functions in advanced manufacturing processes is illustrated in Figure 1.14.

Figure 1.14 Block diagram of adaptive solutions in advanced manufacturing.