建立了基于蚁群算法的激光表面淬火工艺参数神经网络优化系统.用神经网络建立激光表面淬火工艺参数与目标参数的非线性模型,借助蚁群算法搜索决策工艺参数的最优组合,自动优化工艺参数.用VC++6.0开发了激光表面淬火工艺参数优化程序.结果表明,基于蚁群算法的神经网络优化系统用于解决激光表面淬火工艺参数优化问题是可行且有效的.

阐述了半导体激光的特点及其在再制造领域中国内外应用现状、存在问题和发展趋势.半导体激光具有高效、轻量、金属高吸收率等特点,有利于开发节能降耗、高柔性的激光再制造装备与工艺,在再制造产业中有很大的发展潜力.目前,高功率半导体激光器已经逐步走入市场,半导体激光已经在国内外再制造产业中得到应用,以机器人作为执行机构的移动式半导体激光再制造装备与工艺也已经开始应用.半导体激光器的功率还有很大的提升空间,光学整形、光纤耦合等技术还在不断发展,与再制造工艺相关的机器人技术问题也在大力研究,多种材料的熔覆理论与工艺问题不断得到解决,过程检测控制理论技术也在继续攻关.未来的半导体激光再制造技术将突破工件的大小、形状、材质等多方面的制约,迈上一个新的台阶,在更深、更广的空间中应用和发展.

Based on prior calibration with analogist samples, a system, consisting of a CCD camera, a frame grabber and a computer, is developed to in-situ measure the molten pool parameters during laser cladding. The signals captured by CCD vary with change in substrate or clad material, so that, in order to gain the true data of the molten pool parameters, calibration of the instrument must be performed for every set of substrate and clad materials. In this study, a new strategy for on-line acquisition of molten pool parameters is presented. Before measurement of the objective molten pool, cladding is performed to get some analogist samples of the same materials as those used to generate the objective molten pool. The width of a certain segment of the clad bead on the analogist sample is gauged to calibrate the CCD camera-grabbed image of the molten pool which has solidified to this segment of the clad bead. The pool width in the grabbed image can be confirmed to the gauged bead width times the scaling because the width of the clad bead segment must be equal to the width of the related pool. It is assured which type of area in the grabbed image is located on by the pool, and the CCD is calibrated for this set of substrate and clad materials. After calibration, the CCD system is applied to in-situ measure the molten pool parameters during laser cladding. In the presented experiments, the measured values agree well with the actual ones.

Consisting of a CCD camera, a frame grabber and a computer, a system is developed to monitor the molten pool parameters in laser cladding. The molten pool images, grabbed by the CCD camera, are processed by the software, and then the geometry and temperature field distribution of molten pool are measured, which can show the changes in laser cladding. The relationship between the molten pool parameter and the visible defects is revealed, and the visible defects can be detected in-time by diagnosis of the molten pool parameters.

In order to study the processing quality of the laser hardened workpiece,a dynamic detecting system of laser hardening was developed in which a CCD camera was used to capture the thermal radiation images in laser hardening process. The oversaturated phenomenon can be eliminated during the detecting process by selecting the CCD model reasonably.The images of the object were taken by the CCD camera.Then the image data were analyzed by a computer so that image processing made the gray value distribution display clearly,and the hardened band width was obtained by pseudo-color and threshold segmentation processing,which made the software system was developed successfully.The experimental results show that this system can accurately measure the hardened band width,which will contribute to a closed-loop control system for laser hardening for improvement in hardening quality.

Combination of back propagation(BP)neural network and particle swarm optimization(PSO) algorithms is used to optimize process variables during the laser cladding.BP neural network model is developed to express the relationship between the clad process variables and the clad parameters(the width,height of clad bead),and the samples obtained in experiments are used to train network model to form the perfect map relation between input and output.Then,PSO algorithm is used to grabble the suitable values of the process variables.The experimental clad parameters with the process variable values calculated by this optimization method are coincident well with the expected ones.It is verified experimentally that combination of BP neural network and PSO algorithms can help to obtain the expected laser clad quality.

In order to study the size of melt pool in laser cladding,a system for monitoring the melt pool was constructed with a CCD camera and special image-processing software developed on VISUAL C~(++).The edge and geometric parameters of the melt pool were obtained in real-time.The width of the clad layer was obtained in time after calibration correction coefficients of the experimental detection system.Experiment results indicate that the width of melt pool can be detected accurately,and the presented system is useful for online detection of the clad quality.