CN108982096B - Industrial robot crankshaft abrasion detection method based on heuristic rule system - Google Patents

Industrial robot crankshaft abrasion detection method based on heuristic rule system Download PDF

Info

Publication number
CN108982096B
CN108982096B CN201810554894.7A CN201810554894A CN108982096B CN 108982096 B CN108982096 B CN 108982096B CN 201810554894 A CN201810554894 A CN 201810554894A CN 108982096 B CN108982096 B CN 108982096B
Authority
CN
China
Prior art keywords
heuristic rule
rule system
torque
wear
rule
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810554894.7A
Other languages
Chinese (zh)
Other versions
CN108982096A (en
Inventor
徐晓滨
赵状状
翁旭
李建宁
侯平智
黄大荣
王晓兵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN201810554894.7A priority Critical patent/CN108982096B/en
Publication of CN108982096A publication Critical patent/CN108982096A/en
Application granted granted Critical
Publication of CN108982096B publication Critical patent/CN108982096B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/02—Gearings; Transmission mechanisms
    • G01M13/028—Acoustic or vibration analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
    • G01R19/0092—Arrangements for measuring currents or voltages or for indicating presence or sign thereof measuring current only

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention relates to an industrial robot crank shaft abrasion detection method based on a heuristic rule system. The invention uses heuristic rule system modeling to describe the nonlinear mapping relation between the crankshaft torque mean value and torque derivative mean value variables f1 and f2 (input) of the industrial robot and the crankshaft wear degree variable Y (output) of the industrial robot. And optimizing an initial model of a heuristic rule system through limited historical data, and reducing the influence of subjective factors on the model. The heuristic rule system method can calculate the parameters of the model more simply and rapidly, so that the wear degree grade of the crankshaft of the industrial robot can be accurately and rapidly estimated through the heuristic rule system under the condition of setting the characteristics of the torque mean value and the torque derivative mean value. The method improves the estimation precision and the calculation efficiency of the model, and has higher efficiency and advantages.

Description

Industrial robot crankshaft abrasion detection method based on heuristic rule system
Technical Field
The invention relates to an industrial robot crank shaft abrasion detection method based on a heuristic rule system, and belongs to the field of industrial production safety maintenance.
Background
With the improvement of the automation degree and the production line takt, the industrial robot under the high-speed and high-load operation has frequent faults, wherein the fault incidence rate of the wear of the crank shaft of the robot is the highest. Because the crank of the robot crank shaft is used as the input and the output of the second-stage cycloidal gear transmission, after the crank is worn, the crank greatly affects the positioning accuracy of the robot, and simultaneously has the phenomena of abnormal sound, shaking and the like, the service life of the robot is greatly shortened, and even serious safety accidents can be caused.
In the process of detecting the abrasion of the motor cylindrical roller bearing, a student researches that the cylindrical roller bearing to be detected is flatly placed on a special tool clamp by adopting contour detection, finds out the highest point of a roller, measures the contour of a bus of the highest point of the roller, evaluates the height difference between the highest point and the lowest point on the effective sampling length of a bearing area of the roller and judges the abrasion degree according to the height difference; or the bearing roller bus is detected by adopting light transmission detection, namely a linear ruler with higher precision is used for detecting the bearing roller bus, the linear ruler is closely attached to the bearing roller bus in parallel, the light transmission detection is carried out on the gap between the linear ruler and the bearing roller bus, and the abrasion condition of the surface profile of the roller is qualitatively judged according to the light transmission intensity and the light transmission color. The crankshaft morphological structure and the function in the whole system are similar to those of a motor cylindrical roller bearing, so that the motor cylindrical roller bearing abrasion detection based on contour detection and light transmission detection is suitable for crankshaft abrasion detection in principle. However, from the consideration of the efficiency of practical engineering application, the normal and efficient operation of the RV reducer is not influenced when the wear degree of the crankshaft is small, and at the moment, the crankshaft is stopped and disassembled for subsequent wear detection, so that the economical applicability is not realized in engineering. In addition, the wear detection method has higher requirements on the precision and the manufacturing process of the used tool, one set of tool equipment has no universality, a set of special tool clamp and a straight line ruler suitable for crankshaft wear detection is researched and developed, the difficulty is high, the cost is high, the period is long, and the wear detection method has great difficulty in popularization in general companies and factories except professional research laboratories.
In addition to the physical detection methods described above with respect to bearing wear, there are also on-line detection methods of bearing wear based on ultrasonic waves. The principle is that after the pulse of the main controller is sent by the transmitting circuit, the pulse is changed into periodic extremely narrow electric pulse in a very short time, after the pulse is added to the probe, pulse ultrasonic waves can appear on the piezoelectric plate, the ultrasonic waves can be reflected on the upper surface and the lower surface of a workpiece, after reflected waves are converted into electric signals, the time of sound waves transmitted on the upper surface and the lower surface of the workpiece is obtained through the amplifier, the thickness is calculated, and then the thickness is compared to judge the wear degree of the bearing. Although the online detection method overcomes the complexity of a physical monitoring method to a certain extent, a new problem is introduced. It does not take advantage of the many more recognizable and discernable input signals of the crankshaft itself, but rather artificially introduces a master pulse externally. Practice shows that the pulse size selection has great influence on the detected abrasion degree, so that the measurement result is full of uncertainty, and the detection precision is difficult to ensure. In addition, the method can only detect within a certain range of wear degree, and can not effectively and reasonably grade the wear degree of the crankshaft.
Therefore, it is very important to find a crankshaft wear detection method that can use its own signal as input and can precisely determine the fault level while ensuring the accuracy without additionally designing a physical detection tool. Therefore, a heuristic rule system is introduced, the heuristic rule system has the capability of modeling data with incomplete, fuzzy, probabilistic uncertainty, subjective/objective and nonlinear characteristics, the complex nonlinear relation between the input quantity and the output quantity of the controller is described by the heuristic rule system through the expression of knowledge realized by the heuristic rule system and the reasoning of the knowledge based on the evidence reasoning algorithm, then a nonlinear learning model is given, the optimization problem of the initial heuristic rule system parameter given by an expert is solved, the action of a controlled object is controlled by the output of the trained heuristic rule system, and the nonlinear mapping relation between the input variable and the output variable is accurately described.
Disclosure of Invention
The invention designs a detection method based on an heuristic rule system aiming at the uncertainty relation between the torque characteristics of a servo motor of an industrial robot and the wear state of a crankshaft of an RV (rotating vector) speed reducer.
The method comprises the steps that firstly, the input of a diagnosis system can select the mean value of the torque of the motor and the mean value of the reciprocal of the torque, and the output is determined as the wear fault grade of the crankshaft. Secondly, fusing the confidence structure in the item after the activated confidence rule is input by using an Evidence Reasoning (ER) algorithm to obtain the wear level. Finally, the method is verified through actually measured torque data acquired by an industrial robot of a certain model, and the method is proved to have strong superiority.
The invention comprises the following steps:
the method comprises the following steps that (1) a torque signal a1(T) and a current signal b1(T) are acquired by using a vibration sensor and a current sensor which are installed on a servo motor of the industrial robot, wherein the torque unit is mm, the current signal unit is A, a1(T) belongs to the range of-4, 2, b1(T) belongs to the range of-4, the torque signal and the current signal are sampled at the same time every 1 second for T times in total, T < infinityis more than or equal to 1000, and then the sampling time T is 1,2, … and T.
Step (2) averages the signal a1(t) acquired in step (1) as a feature f1(t), calculates the slope of the signal a1(t), and then calculates the average of the slopes as a feature f2 (t).
And (3) outputting a wear fault grade of the crankshaft, and qualitatively dividing the crankshaft into three different wear states: normal state for safe work requirement; moderate wear state with high defective rate; and a heavy wear condition where it does not work properly. The wear failure level is denoted as y (T), and f1(T), f2(T), and y (T) are expressed as vectors P (T) (f 1(T), f2(T), y (T)), and T vectors are obtained in total, and the vector set formed by these vectors is denoted as P (T) | T ═ 1,2, …, T }.
Step (4) establishing rules of a heuristic rule system by using a nonlinear relation between torque mean value and torque derivative mean value variables f1 and f2 (input) and a wear degree variable Y (output), wherein the k-th rule of the heuristic rule system is marked as RkIt is expressed as follows:
Rk:
Figure BDA0001682043980000031
Figure BDA0001682043980000032
Rkis given a rule weight of thetakSatisfies the condition of 0 ≤ thetakLess than or equal to 1; input variable f1And f2The corresponding attribute weights are respectively delta1,δ2And 0 is not less than delta1,δ2≤1；
In the formula (1), the reaction mixture is,
Figure BDA0001682043980000033
and
Figure BDA0001682043980000034
respectively f of input variables of heuristic rule system1And f2Reference value, and has
Figure BDA0001682043980000035
Wherein j is 1,2, QjIs composed of
Figure BDA0001682043980000036
Wherein the elements satisfy
Figure BDA0001682043980000037
xjRepresents the number of values, x, corresponding to the jth input variable reference valuejNot less than 1; are respectively at Q1,Q2Extracting an element as f1、f2Thus combined into a rule, can yield L ═ x in total1×x2The rule is that L is more than or equal to 1, k is 1,2,3, …, and L is the serial number of the rule;
in the formula (1), RkThe latter item attribute is G respectively1,G2,…,GNAnd is combined with LY≤G1<G2<…<GN≤RY，N≥2，
Figure BDA0001682043980000038
λ1,k,λ2,k,…,λN,kAre each G1,G2,…,GNAnd satisfies a confidence value of 0 ≦ λi,k≤1，
Figure BDA0001682043980000039
i＝1,2,…,N；
Wherein, in the formula (1), the initial rule weight is set as thetakInitial attribute weight δ 1j＝1。
Step (5) setting a torque mean value and a torque derivative mean value f1And f2Then, the corresponding crank shaft wear degree estimation results are obtained through a heuristic rule system
Figure BDA00016820439800000310
The method comprises the following specific steps:
step (5-1) setting f1And f2Are each f1 1And
Figure BDA00016820439800000311
superscript 1 represents the input to a heuristic rule system and has
Figure BDA00016820439800000312
Bring them into a heuristic rule system model, calculate the weight of each rule they activate:
Figure BDA00016820439800000313
ωk∈[0,1](3)
in the formula (2), the reaction mixture is,
Figure BDA0001682043980000041
expressed as the jth input variable in the kth rule relative to a reference value
Figure BDA0001682043980000042
Coordinate difference of (c ═ 1,2, …, x)j) The solving method of the coordinate difference value is as follows:
(a) when f isj 1≤Aj,1And
Figure BDA0001682043980000043
when f is presentj 1For Aj,1And
Figure BDA0001682043980000044
coordinate difference of (2)
Figure BDA0001682043980000045
Values are all 1, and coordinate differences of other reference values are all 0;
(b) when A isj,c<fj 1≤Aj,c+1When f is presentj 1For Aj,cAnd Aj,c+1Coordinate difference of (2)
Figure BDA0001682043980000046
The values are given by the formulas (4) and (5) respectively
Figure BDA0001682043980000047
Figure BDA0001682043980000048
At this time, the coordinate difference values of other reference values are all 0;
step (5-2) obtaining an input variable of f1 1And
Figure BDA0001682043980000049
confidence fusion values of different consequent outputs after model inference
Figure BDA00016820439800000410
Figure BDA00016820439800000411
Wherein m isi,k＝ωkλi,k(7)
And λi,kConfidence of the property of the consequent for the activated rule;
step (5-3) obtaining an input variable of f1 1And
Figure BDA00016820439800000412
time crank shaft wear degree estimation result
Figure BDA00016820439800000413
Figure BDA00016820439800000414
The obtained result is the estimated value of the heuristic rule system to the wear degree grade of the crank shaft of the industrial robot.
The key technology of the method is as follows: the method comprises the steps of firstly selecting historical data with certain characteristics to carry out data analysis, obtaining torque signals and current signals collected by a vibration sensor and a current sensor which are installed on a servo motor of the industrial robot, constructing a heuristic rule system which reflects the nonlinear relation between input frequency domain characteristic signals f1 and f2 and an output crankshaft wear degree grade Y, then selecting training samples, and solving the nonlinear optimization problem formed by all parameters in the heuristic rule system model.
The method comprises the steps of firstly establishing a nonlinear mapping relation between torque signals and current signals acquired by a vibration sensor and a current sensor on a servo motor of the industrial robot and the wear degree grade of a crankshaft by utilizing a heuristic rule system. Therefore, the influence of subjective factors is reduced, and the estimation precision and the calculation efficiency of the model are improved.
Drawings
FIG. 1 is a block flow diagram of the method of the present invention.
Fig. 2 is the characteristic data obtained by averaging the acquired time domain vibration signal a1(t), taking the average as the characteristic f1(t), calculating the slope of the vibration signal a1(t), and then calculating the average of the slope as the characteristic f2(t) in the embodiment of the method of the present invention.
FIG. 3 is a graph of crankshaft wear estimates based on a heuristic rule system using training samples in an embodiment of the method of the present invention.
FIG. 4 is a graph of the trace between the estimated and true values of the heuristic rule system model in an embodiment of the method of the present invention.
Detailed Description
An industrial robot crank shaft abrasion detection method based on a heuristic rule system comprises the following steps:
(1) torque signals a1(T) and current signals b1(T) acquired by using a vibration sensor and a current sensor which are installed on a servo motor of the industrial robot, wherein the torque unit is mm, the current signal unit is A, a1(T) belongs to the range of-4, 2, b1(T) belongs to the range of-4, the torque signals and the current signals are simultaneously sampled every 1 second for T times, T < infinityis more than or equal to 1000, and the sampling time T is 1,2, … and T.
(2) When the current signal is stable, averaging the time-domain vibration signal a1(t) obtained in the step (1), wherein the average value is used as a feature f1(t), calculating the slope of the vibration signal a1(t), and then calculating the average value of the slope to be used as a feature f2 (t).
(3) The output is determined as the wear failure grade of the crankshaft, and the crankshaft is qualitatively divided into three different wear states: normal state for safe work requirement; moderate wear state with high defective rate; and a heavy wear condition where it does not work properly. The wear failure level is denoted as y (T), and f1(T), f2(T), and y (T) are expressed as vectors P (T) (f 1(T), f2(T), y (T)), and T vectors are obtained in total, and the vector set formed by these vectors is denoted as P (T) | T ═ 1,2, …, T }.
(4) Establishing rules of a heuristic rule system by using a nonlinear relation between torque mean and torque derivative mean variables f1 and f2 (input) and a wear degree variable Y (output), wherein the k-th rule of the heuristic rule system is marked as RkIt is expressed as follows:
Rk:
Figure BDA0001682043980000051
Figure BDA0001682043980000052
Rkis given a rule weight of thetakSatisfies the condition of 0 ≤ thetakLess than or equal to 1; input variable f1And f2The corresponding attribute weights are respectively delta1,δ2And 0 is not less than delta1,δ2≤1；
In the formula (1), the reaction mixture is,
Figure BDA0001682043980000061
and
Figure BDA0001682043980000062
respectively f of input variables of heuristic rule system1And f2Reference value, and has
Figure BDA0001682043980000063
Wherein j is 1,2, QjIs composed of
Figure BDA0001682043980000064
Wherein the elements satisfy
Figure BDA0001682043980000065
xjRepresents the number of values, x, corresponding to the jth input variable reference valuejNot less than 1; are respectively at Q1,Q2Extracting an element as f1、f2Thus combined into a rule, can yield L ═ x in total1×x2The rule is that L is more than or equal to 1, k is 1,2,3, …, and L is the serial number of the rule;
in the formula (1), RkThe latter item attribute is G respectively1,G2,…,GNAnd is combined with LY≤G1<G2<…<GN≤RY，N≥2，
Figure BDA0001682043980000066
λ1,k,λ2,k,…,λN,kAre each G1,G2,…,GNAnd satisfies a confidence value of 0 ≦ λi,k≤1，
Figure BDA0001682043980000067
i＝1,2,…,N；
Wherein, in the formula (1), the initial rule weight is set as thetakInitial attribute weight δ 1j＝1。
For ease of understanding, it is assumed, by way of example, that
Figure BDA0001682043980000069
Wherein the value of each reference value is A1,1＝7，A1,2＝8，A1,3＝9，
Figure BDA0001682043980000068
Wherein the value of each reference value is A2,1＝0.2，A2,2＝0.3，A3,30.4, provided that Y is satisfied1×f2，G1＝0.4,G2＝1.5,G3＝2.3,G4A total of 2 will yield L9 rules, given the initial θk＝1,δ1＝δ 21, wherein the following rules are partially defined:
R1if f1Is equal to 7 and f20.2, then [ (G)1,0.667),(G2,0.333),(G3,0),(G4,0)]；
R2If f1Is equal to 7 and f20.3, then [ (G)1,0),(G2,0.75),(G3,0.25),(G4,0)]；
R3If f1Is equal to 7 and f20.4, then [ (G)1,0),(G2,0),(G3,0.75),(G4,0.25)]；
……
R7If f1Is equal to 9 and f20.2, then [ (G)1,0.444),(G2,0.556),(G3,0),(G4,0)]；
R8If f1Is equal to 9 and f20.3, then [ (G)1,0),(G2,0.25),(G3,0.75),(G4,0)]；
R9If f1Is equal to 9 and f20.4, then [ (G)1,0),(G2,0),(G3,0),(G4,1)]；
Here, 9 rules are formed, where λi,kThe results from the historical data analysis are analyzed to satisfy the constraints.
(5) Given torque mean and torque derivative mean f1And f2Then, the heuristic rule system acquires the corresponding dataCrankshaft wear degree estimation result
Figure BDA0001682043980000071
The method comprises the following specific steps:
step (5-1) setting f1And f2Are each f1 1And
Figure BDA0001682043980000072
superscript 1 represents the input to a heuristic rule system and has
Figure BDA0001682043980000073
Bring them into a heuristic rule system model, calculate the weight of each rule they activate:
Figure BDA0001682043980000074
ωk∈[0,1](3)
wherein,
in the formula (2), the reaction mixture is,
Figure BDA0001682043980000075
expressed as the jth input variable in the kth rule relative to a reference value
Figure BDA0001682043980000076
Coordinate difference of (c ═ 1,2, …, x)j) The solving method of the coordinate difference value is as follows:
(a) when f isj 1≤Aj,1And
Figure BDA0001682043980000077
when f is presentj 1For Aj,1And
Figure BDA0001682043980000078
coordinate difference of (2)
Figure BDA0001682043980000079
Values are all 1, and coordinate differences of other reference values are all 0;
(b) when A isj,c<fj 1≤Aj,c+1When f is presentj 1For Aj,cAnd Aj,c+1Coordinate difference of (2)
Figure BDA00016820439800000710
The values are given by the formulas (4) and (5) respectively
Figure BDA00016820439800000711
Figure BDA00016820439800000712
At this time, the coordinate difference values of other reference values are all 0;
step (5-2) obtaining an input variable of f1 1And
Figure BDA00016820439800000713
confidence fusion values of different consequent outputs after model inference
Figure BDA00016820439800000714
Figure BDA00016820439800000715
Wherein m isi,k＝ωkλi,k(7)
And λi,kConfidence of the property of the consequent for the activated rule;
step (5-3) obtaining an input variable of f1 1And
Figure BDA00016820439800000716
time crank shaft wear degree estimation result
Figure BDA00016820439800000717
Figure BDA00016820439800000718
The obtained result is the estimated value of the heuristic rule system to the wear degree grade of the crank shaft of the industrial robot.
For the sake of understanding, the model in step (4) is taken as an example, and the model input f is assumed1 17.6 and
Figure BDA0001682043980000081
the actual wear level Y at this time is 1, and equations (2) to (4) are substituted, and the activation rule R can be obtained1、R2、R4And R5The calculated activation weight is: w is a1＝0.28，w2＝0.12，w4＝0.42，w50.18, and 0 for the rest; therefore, the results are taken into the formulas (5) and (6)
Figure BDA0001682043980000082
Brought into formula (7), to obtain
Figure BDA0001682043980000083
The deviation from the true value is 0.083, and the number of rules and the number of the consequent outputs can be increased for improving the precision.
Embodiments of the method of the present invention are described in detail below with reference to the accompanying drawings:
the flow chart of the method of the invention is shown in figure 1, and the core part is as follows: acquiring a torque signal a1(t) and a current signal b1(t) by using a vibration sensor and a current sensor which are arranged on a servo motor of an industrial robot, wherein the unit of torque is mm, the unit of the current signal is A, averaging the acquired torque signal a1(t) (representing a time-domain vibration signal), taking the average value as a characteristic f1(t), calculating the slope of the vibration signal a1(t), then averaging the slope as a characteristic f2(t), and constructing a structure reflecting the input characteristic signal f1And f2And a heuristic rule system for outputting a non-linear relationship between the crankshaft wear failure level Y.
The steps of the method are described in detail by combining a large amount of measured data for analysis, and experimental results prove that compared with an industrial robot crankshaft wear detection method using an MATLAB toolbox fmincon function, the industrial robot crankshaft wear detection method based on a heuristic rule system has the advantage of high calculation efficiency, and the situation that no optimized result exists in the fmincon function is avoided.
1. Torque signals a1(T) and current signals b1(T) acquired by using a vibration sensor and a current sensor which are installed on a servo motor of the industrial robot, wherein the torque unit is mm, the current signal unit is A, a1(T) belongs to the range of-4, 2, b1(T) belongs to the range of-4, the torque signals and the current signals are simultaneously sampled every 1 second for T times, T < infinityis more than or equal to 1000, and the sampling time T is 1,2, … and T.
2. The time domain vibration signal a1(t) obtained in step 1 is averaged to obtain a mean value as a feature f1(t), the vibration signal a1(t) is subjected to slope calculation, and then the mean value of the slope is obtained as a feature f2(t), as shown in fig. 2, wherein a dotted line represents current and a solid line represents torque.
3. The crankshaft is qualitatively divided into three different wear states: normal state for safe work requirement; moderate wear state with high defective rate; and a heavy wear condition where it does not work properly. The wear failure level is denoted as y (T), and f1(T), f2(T), and y (T) are expressed as vectors P (T) (f 1(T), f2(T), y (T)), and T vectors are obtained in total, and the vector set formed by these vectors is denoted as P (T) | T ═ 1,2, …, T }.
4. Establishing a mean signal f reflecting a torque mean signal and a torque slope of the crankshaft1And f2(two-dimensional input to the heuristic rule system) and a crankshaft wear failure degree variable Y (output of the heuristic rule system).
Selecting semantic value of input and output variable, f1The fuzzy semantic value of (a) is described as: minimal (ES 1), very small (VS 1), positive small (PS 1), positive medium (PM 1), positive large (PL 1), very large (VL 1); f. of2The fuzzy semantic value of (a) is described as: minimal (ES 2), minimal (VS 2), positive (PS 2), and median (posiv) small (ES 2)e medium, PM2), positive large (PL 2), very large (VL 2). The reference values are shown in tables 1 to 3:
TABLE 1 f1Semantic value and reference value of
Figure BDA0001682043980000091
TABLE 2 f2Semantic value and reference value of
Figure BDA0001682043980000092
Semantic and reference values of Table 3Y
Figure BDA0001682043980000093
The initial heuristic rule system established is shown in table 4, where the confidence value of the consequent output is given as required based on historical data:
TABLE 4 initial confidence rule base
Figure BDA0001682043980000094
Figure BDA0001682043980000101
5. Combining the step (5) to obtain the given input characteristic f of the training sample1And f2Obtaining the corresponding crank shaft abrasion fault degree estimation results through a heuristic rule system
Figure BDA0001682043980000102
Selecting TN-60 training samples from 2000 groups of vectors, wherein the corresponding input real fault level is Y (t), and the corresponding estimated value of the training samples input through a heuristic rule system is
Figure BDA0001682043980000103
First, the activation weight of the rule is calculated for each input variable. The activation weight omega of each rule in the heuristic rule system can be calculated by using the formulas (2) to (5)k. When t is 320, the input quantity is f1＝0.3369、f20.0062, its pair rule R can be obtained11，R12And R19，R20Respectively is w11＝0.6915,w12＝0.13,w19＝0.1503,w20The activation weight of other rules is 0.0282, that is, 4 rules in the heuristic rule system are activated.
Finally, the activated rules are fused. And fusing the confidence structure of the activated heuristic rule postterms by utilizing an ER algorithm to obtain an output confidence structure about Y. Can be obtained from the equations (5) and (6)
Figure BDA0001682043980000104
For example, consider the above step with respect to f1＝0.3369、f2W 0.0062kAnd initial heuristic rule system consequent confidence lambdaj,kIn the belt-carrying formulas (7) and (8), it is possible to obtain:
Figure BDA0001682043980000111
6. determining heuristic rule system for detecting wear degree of optimized crankshaft, and performing result verification test
According to the parameters in the crankshaft wear degree detection system obtained in the steps, verification is carried out by using sample data which is randomly extracted. Here, it is possible to obtain the estimated curve of the wear level of the crank shaft by the initial heuristic rule system and the estimated curve of the wear level of the crank shaft by the optimized heuristic rule system as shown in FIG. 3, FIG. 3(a) shows an estimated curve of an initial heuristic rule system for a level of crankshaft wear, FIG. 3(b) shows a graph of the estimated crankshaft wear level for the model of the optimized heuristic rule system for the same input variables, where the true data value is denoted by "o", the crankshaft wear level estimate given by the method of the present invention is denoted by "", figure 4 shows a trace between the respective estimates of the wear level of the crankshaft and the true values of the heuristic rule system, wherein "-" represents the curve of the true value of the wear degree grade of the crankshaft, and "-" represents the curve of the heuristic rule system for the estimated value of the wear degree grade of the crankshaft.

Claims (1)

1. An industrial robot crank shaft abrasion detection method based on a heuristic rule system is characterized by comprising the following steps:
acquiring a torque signal a1(T) and a current signal b1(T) by using a vibration sensor and a current sensor which are installed on a servo motor of an industrial robot, wherein the torque unit is mm, the current signal unit is A, a1(T) belongs to the range of-4, 2, b1(T) belongs to the range of-4, 4), the torque signal and the current signal are sampled at the same time every 1 second for T times, and if T is more than or equal to 1000 and less than infinity, the sampling time T is 1,2, … and T;
step (2) averaging the signal a1(t) obtained in step (1) to obtain an average value as a feature f1(t), obtaining a slope c1(t) for the signal a1(t), and then obtaining an average value of the slope c1(t) to obtain a feature f2 (t);
and (3) outputting a wear fault grade of the crankshaft, and qualitatively dividing the crankshaft into three different wear states: the device can be used for a normal state with safety requirements, a moderate wear state with high defective rate and a severe wear state which cannot work normally; recording the wear failure grade as Y (T), and expressing f1(T), f2(T) and Y (T) as vectors P (T) ([ f1(T), f2(T), Y (T)) ], wherein T vectors are obtained in total, and the vector set formed by the vectors is recorded as P (P (T) | T ═ 1,2, … and T };
establishing rules of a heuristic rule system, and adopting a nonlinear relation between torque mean value and torque derivative mean value variables f1 and f2 and a wear degree variable Y, wherein the kth rule of the heuristic rule system is marked as RkIt is expressed as follows:
Figure FDA0001682043970000011
Figure FDA0001682043970000012
Rkis given a rule weight of thetakSatisfies the condition of 0 ≤ thetakLess than or equal to 1; input variable f1And f2The corresponding attribute weights are respectively delta1,δ2And 0 is not less than delta1,δ2≤1；
In the formula (1), the reaction mixture is,
Figure FDA0001682043970000013
and
Figure FDA0001682043970000014
respectively f of input variables of heuristic rule system1And f2Reference value, and has
Figure FDA0001682043970000015
Wherein j is 1,2, QjIs composed of
Figure FDA0001682043970000016
Wherein the elements satisfy
Figure FDA0001682043970000017
xjRepresents the number of values, x, corresponding to the jth input variable reference valuejNot less than 1; are respectively at Q1,Q2Extracting an element as f1、f2Are combined into a rule, which yields L ═ x in total1×x2The rule is that L is more than or equal to 1, k is 1,2,3, …, and L is the serial number of the rule;
in the formula (1), RkThe latter item attribute is G respectively1,G2,…,GNAnd is combined with LY≤G1<G2<…<GN≤RY，N≥2，
Figure FDA0001682043970000021
λ1,k,λ2,k,…,λN,kAre each G1,G2,…,GNAnd satisfies a confidence value of 0 ≦ λi,k≤1，
Figure FDA0001682043970000022
i＝1,2,…,N；
Wherein, in the formula (1), the initial rule weight is set as thetakInitial attribute weight δ 1j＝1；
Step (5) setting a torque mean value and a torque derivative mean value f1And f2Then, the corresponding crank shaft wear degree estimation results are obtained through a heuristic rule system
Figure FDA0001682043970000023
The method comprises the following specific steps:
step (5-1) setting f1And f2Are each f1 1And
Figure FDA00016820439700000217
superscript 1 represents the input to a heuristic rule system and has
Figure FDA0001682043970000024
Bring them into a heuristic rule system model, calculate the weight of each rule they activate:
Figure FDA0001682043970000025
ωk∈[0,1](3)
wherein,
in the formula (2), the reaction mixture is,
Figure FDA0001682043970000026
expressed as the jth input variable in the kth rule relative to a reference value
Figure FDA0001682043970000027
Coordinate difference of (2), coordinate differenceIs solved as follows:
(a) when f isj 1≤Aj,1And
Figure FDA0001682043970000028
when f is presentj 1For Aj,1And
Figure FDA0001682043970000029
coordinate difference of (2)
Figure FDA00016820439700000210
Values are all 1, and coordinate differences of other reference values are all 0;
(b) when A isj,c<fj 1≤Aj,c+1When f is presentj 1For Aj,cAnd Aj,c+1Coordinate difference of (2)
Figure FDA00016820439700000211
The values are given by the formulas (4) and (5) respectively
Figure FDA00016820439700000212
Figure FDA00016820439700000213
At this time, the coordinate difference values of other reference values are all 0;
step (5-2) obtaining an input variable of f1 1And
Figure FDA00016820439700000216
confidence fusion values of different consequent outputs after model inference
Figure FDA00016820439700000214
Figure FDA00016820439700000215
Wherein m isi,k＝ωkλi,k(7)
And λi,kConfidence of the property of the consequent for the activated rule;
step (5-3) obtaining an input variable of f1 1And
Figure FDA0001682043970000033
time crank shaft wear degree estimation result
Figure FDA0001682043970000031
Figure FDA0001682043970000032
The obtained result is the estimated value of the heuristic rule system to the wear degree grade of the crank shaft of the industrial robot.
CN201810554894.7A 2018-06-01 2018-06-01 Industrial robot crankshaft abrasion detection method based on heuristic rule system Active CN108982096B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810554894.7A CN108982096B (en) 2018-06-01 2018-06-01 Industrial robot crankshaft abrasion detection method based on heuristic rule system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810554894.7A CN108982096B (en) 2018-06-01 2018-06-01 Industrial robot crankshaft abrasion detection method based on heuristic rule system

Publications (2)

Publication Number Publication Date
CN108982096A CN108982096A (en) 2018-12-11
CN108982096B true CN108982096B (en) 2020-04-24

Family

ID=64540379

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810554894.7A Active CN108982096B (en) 2018-06-01 2018-06-01 Industrial robot crankshaft abrasion detection method based on heuristic rule system

Country Status (1)

Country Link
CN (1) CN108982096B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109615228B (en) * 2018-12-12 2023-04-07 杭州电子科技大学 Railway transportation risk probability calculation method based on mixed heuristic rule system
CN109932184B (en) * 2019-03-20 2020-08-04 杭州电子科技大学 Marine diesel engine abnormal wear diagnosis method based on union reliability rule reasoning
CN110084518A (en) * 2019-04-29 2019-08-02 杭州电子科技大学 A kind of slope stability evaluation method based on Mixed design confidence rule-based reasoning model
CN110132603B (en) * 2019-05-16 2020-09-29 杭州电子科技大学 Ship diesel engine fault positioning method based on union confidence rule base and ant colony algorithm

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101135601A (en) * 2007-10-18 2008-03-05 北京英华达电力电子工程科技有限公司 Rotating machinery vibrating failure diagnosis device and method
CN102788695A (en) * 2012-07-18 2012-11-21 南京航空航天大学 Identification method of rolling bearing abrasion
RU2013120821A (en) * 2013-05-06 2014-11-20 Александр Александрович Макушин METHOD FOR INDEPENDENT DIAGNOSTICS OF THE DEGREE OF WEAR OF BEARINGS OF INTERNAL COMBUSTION ENGINE ENGINE
WO2014206711A1 (en) * 2013-06-25 2014-12-31 Magna Powertrain Ag & Co Kg Method for monitoring a state of wear of a clutch unit
CN105067467A (en) * 2015-08-26 2015-11-18 北京工业大学 Frictional wear characteristic test device and method for joint curved junction surfaces
CN105139086A (en) * 2015-08-13 2015-12-09 杭州电子科技大学 Track profile irregularity amplitude estimation method employing optimal belief rules based inference
KR101671634B1 (en) * 2015-06-08 2016-11-02 현대로템 주식회사 Method for determining the amount of wear quantity for shaft parts of power transmission system

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101135601A (en) * 2007-10-18 2008-03-05 北京英华达电力电子工程科技有限公司 Rotating machinery vibrating failure diagnosis device and method
CN102788695A (en) * 2012-07-18 2012-11-21 南京航空航天大学 Identification method of rolling bearing abrasion
RU2013120821A (en) * 2013-05-06 2014-11-20 Александр Александрович Макушин METHOD FOR INDEPENDENT DIAGNOSTICS OF THE DEGREE OF WEAR OF BEARINGS OF INTERNAL COMBUSTION ENGINE ENGINE
WO2014206711A1 (en) * 2013-06-25 2014-12-31 Magna Powertrain Ag & Co Kg Method for monitoring a state of wear of a clutch unit
KR101671634B1 (en) * 2015-06-08 2016-11-02 현대로템 주식회사 Method for determining the amount of wear quantity for shaft parts of power transmission system
CN105139086A (en) * 2015-08-13 2015-12-09 杭州电子科技大学 Track profile irregularity amplitude estimation method employing optimal belief rules based inference
CN105067467A (en) * 2015-08-26 2015-11-18 北京工业大学 Frictional wear characteristic test device and method for joint curved junction surfaces

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于知识规则的发动机磨损故障诊断专家系统;陈果 等;《航空动力学报》;20040229;第19卷(第1期);第23-29页 *

Also Published As

Publication number Publication date
CN108982096A (en) 2018-12-11

Similar Documents

Publication Publication Date Title
CN108982096B (en) Industrial robot crankshaft abrasion detection method based on heuristic rule system
Sharma et al. Cutting tool wear estimation for turning
Huang et al. A PNN self-learning tool breakage detection system in end milling operations
CN113305645B (en) Numerical control machine tool cutter residual life prediction method based on hybrid neural model
JP6752866B2 (en) Cutting tool condition inspection system and method
CN102789545A (en) Method for predicating remaining life of turbine engine based on degradation model matching
Kovač et al. A review of machining monitoring systems
Wan et al. Quality monitoring based on dynamic resistance and principal component analysis in small scale resistance spot welding process
CN110570013B (en) Single-station online wave period data prediction diagnosis method
Bittencourt et al. Modeling and identification of wear in a robot joint under temperature uncertainties
CN116991146B (en) Control method and system of ultrasonic cleaning device
TW202026096A (en) Tool life prediction system and method thereof
CN116893643A (en) Intelligent robot driving track safety control system based on data analysis
El Ouafi et al. Artificial neural network-based resistance spot welding quality assessment system
Müller et al. Weld quality characterization by vibration analysis for ultrasonic metal welding processes
Massol et al. An exTS based neuro-fuzzy algorithm for prognostics and tool condition monitoring
Rahmat et al. Accuracy comparison of ARX and ANFIS model of an electro-hydraulic actuator system
Sharma et al. An approach for condition monitoring of a turning tool
Rao et al. Structure fatigue crack length estimation and prediction using ultrasonic wave data based on ensemble linear regression and Paris’s law
Hamrol Process diagnostics as a means of improving the efficiency of quality control
CN114147380B (en) Rotor welding method based on digital twin frame
CN112034048A (en) Beam structure crack positioning method based on multiple frequency response function estimation
CN109359791A (en) A kind of mechanical system degradation trend prediction technique and system
CN104406606B (en) A kind of time-varying window length dynamic Allan variance analysis methods based on fuzzy control
Ferguson et al. A data processing pipeline for prediction of milling machine tool condition from raw sensor data

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant