Condition of the cutting tool in CNC Machining detects in real time

  • Time:
  • Click:78
  • source:LENA CNC Machining
0 foreword are in CNC Machining, of cutting tool condition detect have very important sense, because the attaint of cutting tool affects the quality of treatment and efficiency not only, and cause serious machine tool and person accident possibly still. The attaint of cutting tool has wear away and damaged two kinds of circumstances, wearing away is cutting tool the use up phenomenon of the exterior material that be contacted with workpiece happening in machining a process and chafe and produces; And damaged is cutting tool happening collapse blade, rupture, the phenomenon that model changes to wait and bring about cutting tool to lose cutting ability, it includes brittleness damaged and plasticity damaged again, brittleness damaged is cutting tool is below machinery and concussion action, die in what have not happen to wear away and appear apparently blade, disintegrate, flake etc. And plasticity damaged is cutting tool is when cutting, wait for action as a result of high temperature, high pressure, the phenomenon that plasticity happens to flow on the exterior layer that contacts with work look and loses cutting ability [1] . Current, to cutting tool detect basically collect labour of choose and employ persons to detect, detect from the line and online detect 3 kinds of strategy. Detecting artificially is namely adding man-hour to undertake detecting to the condition of cutting tool by experience by the worker; Detect from the line even if undertake detecting technically to cutting tool before treatment, forecast its life to look to whether be competent to be machined currently; Online detect also say real time detects, the real time in the process that machining namely undertakes detecting to cutting tool, do corresponding processing according to detected result. Current, the algorithm that detects to cutting tool is not little also, the attaint case that some uses the change that accepts stress from place of theoretic computation cutting tool to judge cutting tool [2] [3] , some methods that use sequential analysis undertake detecting to cutting tool [4] [5] , some uses nerve network technology to undertake detecting to cutting tool [6] [7] , return some to use small commutation theory and nerve network technology integratedly to undertake detecting to cutting tool [8] , but they basically are discussed from academic and aspirant travel. In CNC Machining considering cutting tool plasticity damaged is compared scarce, and wearing away to concern to the security of numerical control is not very Dajujike undertakes handling through leaving a line to detect, the ball head cutting tool that the article uses with CNC Machining middling is research target, the real time that to brittleness damaged medium brittleness ruptures detects undertake study, this kinds of happening that rupture, produce the quality to treatment and machine tool itself serious effect. We think cutting tool itself is put in the crackle with small move, use nerve network to build the laden model of ball head cutting tool, through online detect judge this micro-crack to fall to whether can expand in right now laden condition, if the likelihood is patulous, the provides through reducing a knife feed that we think this load is danger will reduce the load that cutting tool place gets, in order to assure the safety of cutting tool. The real time of 1 cutting tool detects   (1) of model of load of ball head cutting tool build the place before be like to narrate, when CNC Machining, the load that cutting tool place gets and very much element are concerned, but the characteristic that considers ball head cutting tool and the need that machine in real time, the article considers to affect a few bigger factors only, namely the deepness of speed of the rotate speed of main shaft, feed, cutting, cutting function that processes data 4 elements, criterion the model of ball head load is F=f(s of   of   of   of   of       , v, h, m) (1)   among them: F -- laden vector; H -- the deepness of cutting; S -- the rotate speed of main shaft; M -- the cutting function of material. V -- feed; Very apparent, type (1) just gave out load and the general relation between each influencing factor, to seek the embody form of load and the relation between each influencing factor, must beg the specific size that gives each elements to be affected to load, for this, perhaps use the mathematical method such as differential geometry to undertake complex derivation, the method that perhaps uses a test reachs the influence coefficient of each elements, but the environment that the model that such building gets used to change hard, the real time that is used in CNC Machining detects the effect is not very ideal. The article uses nerve network the technology handles this model and will the real time that has with Yu Dao detects in. (2) cutting tool detects in real time principle the principle that this cutting tool detects in real time is the cutting depth that measures a cutting tool in real time first and feed the controller of network of nerve of stuff kind input that reachs treatment part with the rotate speed of main shaft has laden consideration, the laden input detector that reach has consideration, quite, if this are laden the crackle below the fatigue requirement that exceeds cutting tool is patulous and laden, reduce the feed rate of cutting tool, and feed speed reduce a quantity to feedback the input information of CNC controller, make CNC controller makes corresponding control, with making laden size changes safe level. This cutting tool detects in real time the principle is shown like plan institute. (3) the structure of the structural nerve network of nerve network has conclusive effect to the character of system of whole nerve network. This load uses the BP structure of three-layer from the nerve network system of adaptive control. According to the analysis above, apparent input layer has 4 node, output layer has 3 node, namely load is in xyz3 directional size. Present problem decides the node of layer of the concealed intermediate is counted namely, the chooses pair of networks study of the division check the number of intermediate concealed layer and computational character have very main effect, it is the key of success or failure of this network structure, if the node of intermediate concealed layer is too little, criterion the network handles complex issue hard, the node of layer of the concealed between Dan Rezhong is overmuch, will make network study time increases quickly, and bring about network study possibly still excessive, make ability of network interference rejection drops. Current, still coach without perfect theory node of intermediate concealed layer is counted choose and just combine actual condition to undertake p reaper chooses to be optimized stage by stage again. Considering this load the character from adaptive control system, we think load is the successive function of feed speed, according to Kolmogorov theorem (successive function table shows theorem) , for theoretic can accurate imitate this successive function, the input layer that is like three-layer nerve network is M node, output layer is N node, criterion intermediate layer should be 2M+1 node. The division check the number that for this we select the concealed layer among should be 2M+1=2 × 4+1=9 node [9] . Accordingly, this nerve network structure is input layer 4 node, node of intermediate layer nine, output a layer 3 node. (4) the function that the main from a of network of line study nerve character of nerve network has study namely, can output a relation according to measuring the input of example certainly namely, the size that self-correcting joins the authority of each node is worth in order to satisfy fixed target. In learning a process, the option that example counts is very important, if sample book is too little, the performance of the network of course study is bad, if example number increases, certainly will increases the study time of the workload that collects example data and network. In the meantime, it is better to be had as a result of nerve network inside insert function and outside insert function poorer, reason example data must include overall pattern and the effect that consider likely random noise. Laden to this nerve network from adaptive control system, its have 4 inputs node, according to the analysis above, we adopt every node to give 4 values, with them different combination regards example as data-in, can get 256 sample book so. Particular way is each input quantities general inside potential span component is become 4 wait for a portion, the method that uses a test measures the laden value that gives to input a circumstance to fall in every kinds. After getting 256 sample book, we are used undertake study from the line, reach the authority between every join node is worth, the nerve network of such course study built model of corresponding cutting tool load, the real time that is cutting tool detects offer a condition. (5) the designs this detector function of detector is to detect is patulous and laden. If the likelihood is patulous, we think this load is dangerous, reduce the load that cutting tool place gets through reducing the feed of cutting tool, in order to assure the safety of cutting tool. For this, we build the mechanical model of cutting tool above all, we simplify the cutting tool in treatment to bear the cantilever a surname of force for end points, and the load that the power that end points place gets gives to use nerve network to beg namely, such, according to material mechanical relevant theory can reach the is cutting tool and machine tool partly union with the biggest stress in cutting tool is in, can beg the stress that gives here. Later, according to rupturing mechanical relevant theory has formulary Da/dn=f(σ , a, c) , among them, a is the length of crackle, n is the frequency of stress, σ is stress, c is the constant that concerns with material [10] . Go up in type, σ , N, C is known quantity or can check through the data, remain to be the length A of function F and crackle certainly. To F, the step that we take is: The micro-crack that assumes cutting tool is in with machine tool union is Ⅰ , Ⅱ , Ⅲ kind the compound crackle of crackle, be in according to this the scale that the size of stress and shearing stress decides these 3 kinds of crackle, such, can establish formula according to of all kinds and specific crackle type. As to A, we are in according to this kinds of cutting tool the average crack length in service life, this method that average length can adopt nondestructive flaw detection undertakes detecting getting to the cutting tool of different operating period. The article put forward 2 conclusion to build pair of CNC Machining through nerve network the method that medium ball head cutting tool detects in real time, this method can be real time the bulk that gets load to place of cutting tool of head of the ball in treatment has consideration, the laden level that judges this load to whether exceed cutting tool to expand in the crackle below stress exhaustion condition through detecting in real time makes corresponding processing. This method undertook simplifying reasonably to affecting the factor of load, make this algorithmic efficiency that controls a model very tall, because this is special,fit the need that detects in real time. Although the article is research target with the ball head cutting tool in numerical control, actually, the principle of this method also can be used in other treatment and other cutting tool, for instance, electric treatment. CNC Milling