Elsevier

Journal of Manufacturing Processes

Volume 57, September 2020, Pages 222-232
Journal of Manufacturing Processes

Precise control of variable-height laser metal deposition using a height memory strategy

https://doi.org/10.1016/j.jmapro.2020.05.026Get rights and content

Highlights

  • An IBPF cladding nozzle was applied for deposition of the overhanging structures.

  • A deposition height sensor and a closed-loop control system were developed.

  • The uneven top surface produced during the variable height deposition process can be memorized and smoothed by the PI controller.

  • Segmented deposition heights control strategy was proposed for the deposition of unequal-height parts.

Abstract

In respect of laser metal deposition (LMD) 3-D additive manufacturing technology, the method of non-parallel slicing is applied for the deposition of the overhanging and unequal-height parts and is capable of improving the deposition accuracy, stability and efficiency. Nevertheless, the produced uneven top surface derived from variable height deposition process has a possibility to result in instability and even failure of the deposition. In this study, the novel control strategies for the precise control of the unequal-height part were proposed. An inside-beam powder feeding nozzle was applied for the deposition with large overhanging angle. Based on a developed deposition height control system, multiple PI controllers were designed for varying desired deposition heights in an unequal-height layer. The actual heights of the uneven top surface in different positions were memorized by the deposition of a layer, and the process parameters for the next layer could be planned in advance to smooth this unevenness. Under the proposed control strategies, a geometrical stable and convergent process was made achievable when depositing a “fan-shaped” part and a “bent-pipe-shaped” part as the typical overhanging and unequal-height structures. The microstructures were observed to be uniform and compact, and the grain sizes and microhardness in different height positions were found to vary slightly.

Introduction

The laser metal deposition (LMD) is a freeform additive manufacturing technology, which takes the laser beam as the heat source and applies synchronous powder/wire feeding of metallic materials. The LMD for 3D forming and repairing is widely used in spaceflight and aircraft apparatus, machine manufacturing, automobile and shipbuilding industry. Many institutes and enterprises are attaching vital importance on this technology [[1], [2], [3], [4]]. Compared with the selective laser melting (SLM) technology [5,6], the LMD is advantageous in unlimited forming dimension, relatively low costing, compact microstructure, and the deposition of functionally gradient materials [4]. However, the shape accuracy control of the LMD is far more difficult than the SLM. The LMD can only achieve rough forming with a simple shape at present.

Plenty of machine component parts and important structures in engineering and equipment have exhibited the characteristics of overhanging and unequal-height structure, like bent pipe structures in automobile engine, twisted compressor blades, and so on. Nevertheless, a majority of the present LMD path-planning strategies/software for 3D models use parallel layering, equal-height layers, invariant laser spot size, and invariant deposition direction. These fixed process parameters restrict the LMD technology from a complete free forming. Examples are the deposition of such basic overhanging structures as a curved part (Fig. 1a), a “fan-shaped” part (Fig. 1b) and a “bent pipe-shaped” part (Fig. 1c). If a traditional parallel slicing method is applied, the overhanging angle will increase from bottom up, until there is no substrate below for the support and then the molten pool collapses. Many research institutes proposed the method of non-parallel slicing that adapts the curvature change direction of the CAD model, and the deposition keeps tangential towards the growth direction of the curved part [7,8]. A most applied implementation method is to deflect the base and the substrate using multi-axis machine tool. Boisselier et al. [9] developed a 5-axis machine tool with inert atmosphere chamber for the DMD process and they deposited the parts with overhanging structures. DMG Mori Co., Ltd [10]. devised an additive-subtractive machine tool. The high shape accuracy can be achieved and the deposition efficiency is 20 times more efficient than the SLM process. Lee and Jee [11] suggested a methodical approach to multi-axis slicing algorithms for the deposition of overhanging features on a metal part. Our research group proposed a method of deflecting tool. The spatial orientation of the laser cladding nozzle can be varied to deposit structures with a large overhanging angle [12]. Dwivedi et al. [13] suggested a mathematical model for a deposition process planning of slender structures by deflecting the coaxial nozzle. The principles of the abovementioned approaches were mostly the deposition of asymmetrical cross-section tracks with invariable deposition height (Fig. 1a) and invariable process parameters. However, the variable-height deposition of the feature structures shown in Fig. 1b and c will be problematic when the traditional invariant-parameter method is applied.

The stacking of multiple parallel tracks (Fig. 2a) with varying lengths is a commonly applied route planning method for deposition of unequal-height structures at present. Hu et al. [14] put forward a sub-slicing method. The first several layers were deposited with parallel layering, which produced a “staircase-shaped” surface, and then an inclined layer was deposited to cover and smooth this uneven surface.

In order to eliminate the “staircase effect”, decrease the overall production time and reduce the repeated heating, some researches proposed the non-uniform slicing method (Figs. 1b, c and 2b). The biggest challenge of this method lies in the deposition strategy for the unequal-height layer. Singh and Dutta [15] presented a task framework for the multi-direction slicing of CAD-models, including the route planning of non-uniform layers. Toyserkani et al. [16] suggested two experimental based model structures. The variation of the clad height could achieve an excellent consistency with the model prediction. Zhang and Liou [17] developed an adaptive slicing algorithm for the deposition of nonuniform thickness layer, but the achievement still needs the aid of machining process. Ruan et al. [18] suggested an unequal-thickness slicing strategy. With the assistance of an empirical model, the designed deposition height with a variable value was controlled by varying the scanning speed. A sine curve shaped part with a varying deposition height was formed. Panchagnula and Simhambhatla [19] deposited varying height layers by varying the wire feed speed or the torch speed based on a regression model for prediction of the bead geometry. However, the top surface of each deposited variable-height layer was not flat, for which additional face milling operation is needed. Wang et al. [20] varied the powder feeding rate in real-time from 0.5 g/min to 3 g/min with experience, and deposited a cladding track with a varying deposition height from 0.05 mm to 0.46 mm. Wang et al. [21] increased the Z-increment of the nozzle from 0.1 mm to 0.2 mm to deposit oblique thin-walled part. Zhao et al. [22] varied the process parameters in different unit blocks for the deposition of a layer with uneven thickness. Prakash et al. [23] proposed a process-adaptive variable slice thickness strategy though an empirical approach. In our previous study, Zhou et al. [24] deposited unequal height tracks with empirically determined variable scanning speeds. Nevertheless, this open-loop method could encounter difficultly in ensuring that all the different deposition heights reach the different designed heights, which was easy to cause uneven surface and collapse of the molten pool. The deposited “fan-shaped” part exhibited a relative low surface accuracy and some defects in the middle part. The deposition stability and repeatability were weak. The review article [1] mentioned that the current non-parallel layer slicing strategies rely on empirical models to predict process parameters in advance in order to achieve varying layer heights. However, the predicted process parameters with continuously varying values cannot guarantee a geometrical stable deposition process, and the deposition heights can hardly reach the desired heights at different positions.

In order to acquire higher deposition performance and accuracy, a large number of electronic monitoring and control methods were studied to support the LMD process [25,26]. Some closed-loop control strategies were developed for the variable height deposition. Hua and Choi [27] used two optical deposition height sensors and designed a fuzzy-logic based controller. The height was controlled by varying the laser power. Samples with a ramp geometry were deposited with a varying height from 0.14 mm to 0.44 mm, which corresponds to the variable laser power from 250 W to 1600 W. Fathi et al. constructed a feedback control system to control the deposition height, and two kinds of sine-shaped variable-height parts were deposited using a designed PID controller [28] and a PID with sliding mode controller [29] with excellent control performances.

The aforementioned variable height deposition strategies using open-loop and closed-loop control methods remain incapable to well achieve a geometrical stable and accurate deposition process, and the sample parts were deposited with only a small number of layers. Compared with the uniform deposition, the continuously varying process parameters in different deposition heights is more likely to produce an uneven top surface, as illustrated in Fig. 3. A small unevenness can increase after the next several layers of deposition. Therefore, the control task is not only to realize a variable height track in every single layer, but also to measure and compensate the produced convex and concave positions.

The PID controller for the deposition height presented in reference [28] demonstrated remarkable dynamic performances, and the controller had compensated two slots in the substrate. However, the molten pool moved about 2 mm during the settling time, and this 2 mm area in the forming part was not smoothed and would lead to a new unevenness. Zeinali and Khajepour [30,31] built an adaptive fuzzy inverse dynamic model for the LMD process and devised a sliding mode controller, which is capable to compensate unknown and uncertain disturbances like slots with excellent control performance. However, the settling time and overshoot remained.

Besides the unevenness caused by the settling time and overshoot during the closed-loop control process, the consumed time in the feedback process is worthy of consideration. As shown in Fig. 3, a convex point is measured at the time point t1. After the time delay of the feedback process in the hardware, the cladding nozzle has moved with a scanning speed v to another height position at the time point t2, and then the execution of the actuators ceases to have effect on the previous height point. This controlled value, as well as the measured value shifts with the movement of the molten pool. Furthermore, the high frequency change in the control input value requires a high real-time performance of the control system and is possible to cause a geometrical unstable deposition process and non-uniform heat absorption.

In order to address the problems as mentioned above, the following methods were proposed in this study for the design of a new measurement and control strategy:

  • (1)

    In each layer of the deposition, a non-uniform single track with a variable height was deposited in one-step with a single pass scanning, as shown in Fig. 2b. A closed-loop control strategy was devised;

  • (2)

    The measured height values in different positions on a cladding track can be memorized after the deposition of a layer, and the scanning strategy and process parameters for the next layer can be planned in advance. The uneven top surface is then smoothed after multi-layers.

Section snippets

The principle of inside-beam powder feeding nozzle

In order to maintain the nozzle direction normal to the current layer shown in Fig. 1, our research group developed an inside-beam powder feeding (IBPF) nozzle [32]. The principle and product of this nozzle are illustrated in Fig. 4a and b respectively. Compared with the coaxial powder feeding technology [33,34], which is mostly studied and commonly applied in the market at present, the IBPF nozzle splits the solid laser beam into an annular and cone-shaped focusing laser beam. If the annular

The “self-healing” effect

The “self-healing” effect in the LMD process was validated in multiple studies [36,37]. In a certain range of the defocus distance d (Fig. 4a), the deposition height of the actual layer varies with the working distance w (Fig. 4a) inversely. That means in Fig. 3, the deposition height of a new layer is lower in convex points and higher in concave points with a varying working distance. Therefore, the uneven surface can be automatically smoothed after several layers under an open-loop control.

The principle of multiple PI controllers with the height memory strategy

As shown in Fig. 7, an unequal-height and equal-width cladding track in a layer can be split into a number of segments, and the designed (desired) heights of each segment are different. Let j be the segment number and the desired height values hd(j) for each layer are different. The process parameters for each segment must be different as well. Based on the PI control strategy designed in Section 3, the deposition heights of the segments are subjected to control respectively. Each segment j has

Conclusions

In this study, an investigation was conducted into the precise control strategies for the deposition of the unequal-height parts in the LMD technology. The important methods and results are summarized as follows:

  • (1)

    In order to deposit the overhanging structure, which consists of unequal-height layers, an inside-beam powder feeding nozzle was applied. The surrounding shielding/alignment gas can restrain the single powder/gas flow from divergence. If the spraying direction is inclined to a large

Acknowledgements

The authors acknowledge the financial support of the National Natural Science Foundation of China (Grant No. 61903268), the Natural Science Foundation of Jiangsu Province (Grant No. BK20190823) and the China Postdoctoral Science Foundation Grant (Grant No. 2019M661921).

References (39)

  • S. Shi et al.

    Study of cobalt-free, Fe-based alloy powder used for sealing surfaces of nuclear valves by laser cladding

    Nucl Eng Des

    (2012)
  • W.E. Frazier

    Metal additive manufacturing: a review

    J Mater Eng Perform

    (2014)
  • D. Gu et al.

    Laser additive manufacturing of metallic components: materials, processes and mechanisms

    Int Mat Rev

    (2013)
  • B. Vayre et al.

    Metallic additive manufacturing: state-of-the-art review and prospects

    Mech Ind

    (2012)
  • J. Xu et al.

    A review of slicing methods for directed energy deposition based additive manufacturing

    Rapid Prototyping J

    (2018)
  • LASERTEC 65 3D

    (2014)
  • K. Lee et al.

    Slicing algorithms for multi-axis 3-D metal printing of overhangs

    J Mech Sci Tech

    (2015)
  • B. Hu et al.

    Layered-deposition manufacturing complex metal parts by unequal thickness slicing

    J Huazhong Univ Sci Technol (Nat Sci Ed)

    (2011)
  • P. Singh et al.

    Multi-direction slicing for layered manufacturing

    J Comput Inf Sci Eng

    (2001)
  • Cited by (0)

    View full text