With the construction of automated docks, health monitoring technology as a parallel safety assurance technology for unmanned hoisting machinery has become a hot spot in the development of the industry. Hoisting machinery has a huge structure and numerous welded joints. The complexity and nonlinearity of the welded structure itself makes the structural failure parts random and difficult to arrange for monitoring sensors. In order to solve the problem of effectiveness and stability of the sensor arrangement method for monitoring the structure of hoisting machinery. Using the global and local search capabilities enhanced by the complementary search mechanism, a structural vibration monitoring sensor placement algorithm based on the harmony genetic algorithm is proposed. Firstly, the model is established for modal analysis to obtain the displacement matrix of each mode. Secondly, the optimal parameter combination is established through parameter comparison, and the random search mechanism is used to quickly search in the modal matrix to obtain the preliminary solution, and then the preliminary solution is genetically summed The mutation operation obtains the optimized solution, and the optimal solution is retained through repeated iterations to realize the decision of the vibration sensor layout of the crane structure monitoring. Combining the comparison test of harmony genetic algorithm, harmony search algorithm and genetic algorithm, the fitness of harmony genetic algorithm in

Hoisting machinery is an important infrastructure in the construction of the national economy, an important equipment in the construction of major national projects, and also a pioneer in the implementation of major national strategic plans. As of the end of 2020, my country has registered 2.5384 million units of cranes in use, accounting for 15.40% of the total amount of special equipment. The development of my country’s lifting machinery industry started late and its design concepts are backward. The lifting machinery industry has congenital “inadequate design” and acquired “maintenance disorder” and other outstanding problems. In 2020, a total of 107 special equipment accidents and related accidents occurred nationwide, with 106 deaths, of which 27 were crane accidents, accounting for 25.23%, and 31 deaths, accounting for 29.25%. Crane accidents have been ranked as special equipment for many consecutive years. At the forefront of equipment accidents [

Sensor optimal placement methods and evaluation criteria are affected by multiple factors such as the complexity of the analyzed object’s structure, the number of nodes, the number of units, the order of structural vibration response, and the number of expected sensors to be deployed, and each has its own characteristics. Representative methods include Fisher information matrix method [

With the development of intelligent technology, the methods of intelligent optimization algorithm, multi-dimensional sensor parameter fusion algorithm and multi-optimization algorithm cross fusion have been used for sensor arrangement optimization. For example, the Euclidean distance derived from analytic geometry by Cao et al. [

In the traditional harmony search algorithm, each tonal variable in a group of harmony is the value of the independent variable corresponding to the problem. However, in the optimization problem of measuring point layout based on the modal matrix, the harmony needs to be converted into measuring points. Harmony elements are encoded. When the number of selectable measuring points in the system is n, the length of the harmony is n, where each element in the harmony represents the decision variable of the corresponding coded measuring point. When the element value is 1, it means that the measuring point is selected, and vice versa 0 means it is not selected. HGSA combines the ideas of harmony search algorithm and genetic algorithm. On the basis of the harmony search algorithm quickly obtains the local optimal solution, the genetic algorithm introduces the idea of survival of the fittest. Through selection operation, crossover operation, mutation operation, etc., first through selection and crossover, So that high-quality genes (elements) are retained and inherited to the next generation (the next set of harmony), and then based on a certain genetic mutation probability, new genes are obtained through genetic mutation to form a new next generation, increasing genetic diversity (overall Search ability) to get a better solution.

According to the geometric dimensions of the model to be analyzed, a finite element model is established, the mesh is divided, the boundary conditions of the structure are defined, and the modal analysis is performed. Enter the finite element post-processing to obtain the m-order modal results of the structure, extract the n node numbers and the x-direction, y-direction, z-direction displacements

In the formula, _{nm} represents the displacement of the n-th node in the m-th order mode.

Randomly extract j groups of node combinations from the full set solution space as the harmony memory bank, which is shown in

In the formula, s represents the number of sensors, and represents the displacement of the m-th order mode corresponding to the randomly selected node.

Randomly generate a variable _{1}_{1}_{1}_{1}

Randomly generate a variable _{2}_{2}_{2}

Modal Assurance Criterion (_{ij} = 1(_{ij} = 0(_{new}_{worst}_{new}_{worst}_{worst}

In the formula, φ_{i} represents the _{j} represents the _{i} and φ_{j} are column vectors. The values of

Using the idea of roulette selection algorithm, select the appropriate selection operator according to

The two parents determined have a certain probability (crossover probability

The offspring individuals have a certain probability (variation probability

Cycles in turn until the specified number of iterations _{max}

Taking the single main beam hoisting machinery as the analysis object, the overall length of the main beam structure is 2000 mm, the width is 190 mm, and the height is 230 mm. The thickness of the steel plate is 6 mm. The material is Q345 steel, and the overall connection method is welding. The three-dimensional model of the hoisting machinery is shown in

The modal deformation cloud diagram of each order of the single main beam hoisting machinery is shown in

Mode | Frequency(Hz) | Deviation | |
---|---|---|---|

Solid | Shell | ||

First-order mode | 199.21 | 205.69 | 0.033 |

Second-order mode | 222.48 | 228.62 | 0.028 |

Third-order mode | 515.36 | 520.5 | 0.010 |

Fourth-order mode | 562.66 | 573.08 | 0.019 |

Fifth-order mode | 577.88 | 594.92 | 0.029 |

Sixth-order mode | 767.78 | 797.14 | 0.038 |

Mode | Displacement (mm) | Deviation | |
---|---|---|---|

Solid | Shell | ||

First-order mode | 4.69 | 4.645 | 0.010 |

Second-order mode | 4.9313 | 4.877 | 0.011 |

Third-order mode | 6.0117 | 5.866 | 0.024 |

Fourth-order mode | 5.3864 | 5.418 | 0.006 |

Fifth-order mode | 4.9669 | 5.0301 | 0.013 |

Sixth-order mode | 4.654 | 4.396 | −0.055 |

By comparing the analysis parameters of the solid model and the shell element model in

Harmony genetic algorithm selection probability

Based on the comparative analysis of

Based on the comparative analysis in

Through optimization and comparison, the final selection of initial parameters

The first 6 modes of the main girder structure of the hoisting machinery are isolated and the sensor placement of each mode has no effect on the observation of other modes, so the number of sensors is 6 [

Using the HGSA algorithm, there are 6 sensors, HMCR of 0.7, PAR of 0.4, CRO of 0.9, VAR of 0.6, and 3000 search iterations in the

Direction | |||
---|---|---|---|

point 1 | 2531 | 5850 | 7887 |

point 2 | 8499 | 3423 | 4032 |

point 3 | 8508 | 2310 | 6685 |

point 4 | 9423 | 94 | 7158 |

point 5 | 2001 | 7160 | 6395 |

point 6 | 4492 | 3921 | 2572 |

optimal value | 0.0045 | 0.0084 | 0.0058 |

Using the HS algorithm, the number of sensors is 6, the initial harmony memory is 30, the

Direction | |||
---|---|---|---|

point 1 | 6577 | 888 | 1656 |

point 2 | 2526 | 6821 | 3701 |

point 3 | 4992 | 1196 | 2363 |

point 4 | 292 | 749 | 7541 |

point 5 | 9643 | 8617 | 9610 |

point 6 | 1063 | 7266 | 7525 |

optimal value | 0.1788 | 0.1186 | 0.2052 |

Using GA algorithm, there are 6 sensors,

Direction | |||
---|---|---|---|

point 1 | 2829 | 3024 | 1440 |

point 2 | 9193 | 9694 | 2064 |

point 3 | 5508 | 1584 | 2506 |

point 4 | 9230 | 5836 | 8037 |

point 5 | 4596 | 992 | 3914 |

point 6 | 971 | 6571 | 7461 |

optimal value | 0.0096 | 0.0122 | 0.0148 |

For single girder hoisting machinery, the HGSA algorithm, HS algorithm and GA algorithm are used to search for sensor placement. The analysis and convergence of the three algorithms is shown in

The comparative analysis shown in

It can be seen from

The three search algorithms are inevitably with random characteristics, and repeated runs may get different local optimal solutions. Therefore, 30 times of analysis are used to compare the maximum (Best), minimum (Worst), average (Avg), standard deviation (Std) and average probability of deviation from the optimal path (Error) of their fitness respectively, and deviate from the optimal path. The average probability of is shown in

In the formula, _{i}) is the fitness value obtained from the _{0}) is the fitness value obtained as expected. Therefore, the parameters of the three algorithm methods HGSA, HS and GA are shown in

Fitness | GA | HS | HGSA |
---|---|---|---|

Best | 0.0096 | 0.1788 | 0.0045 |

Worst | 0.1108 | 0.2941 | 0.0630 |

Avg | 0.0347 | 0.2391 | 0.0247 |

Std | 0.0238 | 0.0270 | 0.0153 |

Error | 38.77% | 856.43% | 1.10% |

Fitness | GA | HS | HGSA |
---|---|---|---|

Best | 0.0122 | 0.1186 | 0.0084 |

Worst | 0.0878 | 0.2889 | 0.0450 |

Avg | 0.0293 | 0.2228 | 0.0202 |

Std | 0.0166 | 0.0448 | 0.0085 |

Error | 17.18% | 791.23% | 19.34% |

Fitness | GA | HS | HGSA |
---|---|---|---|

Best | 0.0148 | 0.2052 | 0.0058 |

Worst | 0.1031 | 0.3915 | 0.0918 |

Avg | 0.0542 | 0.3135 | 0.0386 |

Std | 0.0272 | 0.0418 | 0.0245 |

Error | 116.83% | 1154.02% | 54.43% |

Through 30 calculation tests, the optimal fitness value obtained by the HGSA algorithm in the

HGSA algorithm is better than HS algorithm and GA algorithm in both the optimal fitness value and the worst fitness value. The average value and standard deviation of 30 tests are also better than HS algorithm and GA algorithm. The average probability of HGSA deviating from the optimal path The smallest. It shows that HGSA has stronger ability to reach the optimal value, has better stability, and is better in avoiding search randomness.

Hoisting machinery has a huge structure, variable cross-section, and many welded joints. Whether the sensor arrangement can cover the most critical or crack-sensitive area of the structure is a key issue for the implementation of the lifting machinery health monitoring technology and method.

In order to solve the problem that the search algorithm for the vibration sensor layout of the main beam of the crane is easy to fall into the local optimum, a HGSA is proposed that has the advantages of simple and fast harmony search algorithm and strong global search ability of genetic algorithm., The use of complementary search mechanism to enhance the algorithm’s global search ability and local search ability, forming a hoisting machinery structure monitoring vibration sensor arrangement method.

In view of the large number of nodes in the solid element model, the obtained node combination may contain the internal nodes of the element, and the sensor cannot be installed in the actual project. The shell element is used to replace the solid element for modal analysis to obtain the modal matrix of the sensor arrangement. Algorithm application provides support. Compared with the solid element, the shell element has the same eigenfrequency of each order, and the shell element has a smaller number of nodes than the solid element, and has a smaller modal matrix dimension, which greatly reduces the search operation of the modal matrix in the sensor placement decision. quantity.

Through parameter comparison and analysis, the optimal HGSA initialization parameters are determined. Taking the main beam of the hoisting machinery as the object, the HGSA, HS and GA algorithms are studied in the vibration sensor placement decision, and comparative analysis is carried out. The test results show that while ensuring the efficiency of the solution, HGSA has a stronger search ability, better stability, and is better in avoiding search randomness.