Cutting-Tools Degradation Assessment for Structural-Steel Machining Centers

High-rise buildings with steel skeleton are constructed with cut and drilled structural steel. A structural-steel machining center is composed of bandsaw machines, drill machines and conveyors. The monitoring of cutting tools such as saw bands and drills may prevent unexpected broken tools and too worn tools to machine the workpieces. These interruptions yield extra scraps, cost increase especially for precious materials and delivery delay, and sometimes cause online technicians hazardous. Therefore, for the cutting tools it is demanding to effectively monitor the conditions and even estimate their remaining useful life based on reliable characterization of measured signals during machining; that is, to foresee the time point at which the tools need to be replaced. In industrial practice, non-destructive detection methods such as vibration, acoustics, current and temperature sensing have been often used as indicators for judging tool wear and faults. The acquired sensor signals are used to monitor and assess the health conditions of cutting tools through appropriate signal analysis methods for extracting fault features. Recently, the techniques regarding artificial intelligence and machine learning developed for condition monitoring and prognostics of cutting tools has drawn considerable attention. In the study extracted features and self-organizing mapping were employed to assess the health status of the saw bands and drills of a structural-steel machining center.

The self-organizing maps (SOM) algorithm, unsupervised learning neural network, was proposed by Kohonen in the 1980s and has been applied to different fields, including speech recognition, image processing and the diagnosis of mechanical components as well. It is a kind of neural network technology. “Self-organizing” refers to the ability to learn and organize information without giving corresponding labels. This uses benchmark data to train the SOM structure. When there is new input, the best matching unit (BMU) will appears on the map. The distance between the new input data and the BMU can be used to evaluate the degradation state of the tool. A larger distance value means that the new input data is not similar to the trained baseline, and a smaller distance value indicates that the new input data is close or similar to the baseline.

Our experimental bench, the structural-steel machining center, comprises a bandsaw machine, a drill machine and a conveyer. The gantry-type horizontal bandsaw machine (CNC1100LDN, Cosen Mechatronics Co., Taiwan) has a weight of 10287 Kg and the dimension 5300´5000´3400 mm, and the three-axis drilling machine (D11, Cosen Mechatronics Co., Taiwan) weights 14200 Kg with the dimension 6074´2408´4225. Two accelerometers were mounted on the left and right saw-band holders of bandsaw machine, measuring the vibration in the saw-band traveling and lateral directions, respectively. A tri-axial accelerometer was mounted on the vertical drilling shaft to measure the axial and two radial vibration; besides, a current clamp probe was used to pick up motor current.

From the accelerometer, it is found that the left arm travel direction and the right arm lateral direction are most effective for measuring the sawing signal. From the preliminary spectrum analysis, it is observed that there is an extreme value at the frequency around 600 Hz and twice as long from the new saw band and the old saw band. Therefore, the peak value frequency, the amplitude corresponding to the peak value frequency, and the RMS value of 2.5 to 5.5 times harmonics frequency are selected as the feature, and then self-organizing map algorithm is performed to obtain the distribution status of the sawing band in the prior, middle and last period, as well as the fulltime status trend of the saw band. The message can help the maintenance activities of the machine, reducing downtime, improve production efficiency, and further predict the decline of tool performance to more effectively using the remaining useful life of the tool before maintenance and downtime in the future.