With the continuous development of the times, the hydraulic system of construction machinery has become one of the important components of the hydraulic system and has been widely used in the construction machinery industry. If the hydraulic pressure of the construction machinery detects a fault, the entire production line will stop working, resulting in heavy losses to the enterprise. The hydraulic system of construction machinery is expensive in terms of its own cost and maintenance cost. Therefore, the control detection and fault diagnosis of hydraulic system of construction machinery has the highest value in engineering applications. In this paper, the neural network model used is based on the classification characteristics of neural network in machine learning. Through extracting and standardizing the sample data processing function, the hydraulic control detection and fault diagnosis of construction machinery have been successfully realized. For the defects of parameter selection in machine learning, such as the difficulty of selecting parameter attributes, and the difficulty of falling into local optimization in the face of complex structure, the diagnosis method based on multi-layer hidden MLP model is adopted. This method can directly extract the feature quantity and input it into the MLP model for proper training, so as to make the learning network more in-depth, so as to ensure the reliability of production and improve the accuracy of model fault diagnosis, and use the neural network diagnosis method to process multi-dimensional data in machine learning.
Research Article
Method of hydraulic control detection and fault diagnosis of construction machinery based on machine learning
https://doi.org/10.21203/rs.3.rs-2729902/v1
This work is licensed under a CC BY 4.0 License
Journal Publication
published 08 Jun, 2023
You are reading this latest preprint version
machine learning
Construction machinery
Hydraulic control detection
failure diagnosis method
The level of science and technology with the development of The Times presents a rapidly rising trend, this trend effectively improves the construction machinery hydraulic control detection and fault diagnosis technology development speed, provides a strong guarantee for the operation of the hydraulic system [1]. Although the diagnostic error detection of hydraulic control technology in construction machinery has made great progress, if there is a power failure, it will lead to the failure of the power hydraulic system, as long as the power source problems, the consequences will be catastrophic. In order to prevent and prevent the occurrence of accidents, it is urgent to vigorously promote the deartification of fault diagnosis technology [2]. The deartification of fault diagnosis technology must be advanced, scientific and efficient intelligent misdiagnosis technology. After decades of development, the technology has obtained obvious economic benefits, and many companies will introduce a large number of talents in this field. Powered by new technologies such as the Internet, neural networks and learning machines, the next generation of machine learning has made significant progress [3]. In some areas, machine learning will surpass human level in the future, and we can see that artificial intelligence will be applied to all walks of life in the near future to become the intelligence of everything, from intelligence to intelligence, towards human-machine integration [4–5].
It is pointed out in the literature that the one-dimensional convolutional neural network model is constructed and applied to the hydraulic control detection and fault diagnosis of construction machinery, and the application of machine learning in the one-dimensional convolutional neural network is introduced [6]. Compared with convolutional neural network, BP neural network and multi-hidden layer multi-layer perceptual model do not need manual extraction function, nor need normalization from the data, but through layer training and layer assembly automatic extraction function, and perform fault diagnosis. The literature shows that in many problems of fault diagnosis, it is difficult to construct a high-precision mathematical model if the structure of the equipment to be detected is too complex [7]. At the same time, it does not better represent the knowledge of experts in the relevant field, so the historical data of the device becomes an important means to understand the performance degradation of the device. The basic idea of using machine learning to diagnose errors is to study the historical data of the device under certain functional limitations, to capture the mapping mechanism contained in the data, and to obtain some relationships between mathematical models, and even to use this model for fault diagnosis [8]. Literature points out that at present, intelligent fault diagnosis of construction machinery hydraulic is mainly based on sample data. Obtain as much error data as possible from construction machinery hydraulic to form a database, and then diagnose faults in the following ways: fault matching. The characteristic signal of hydraulic engineering machinery is detected in real time, and the diagnosis algorithm is used to analyze the fault diagnosis of hydraulic control detection and fault diagnosis of construction machinery [9]. The above process is mainly based on machine learning to realize hydraulic control detection and fault diagnosis of construction machinery. The literature shows that fault tree analysis is one of the methods to evaluate the reliability of the system, and there are many problems for the construction machinery hydraulic system with complex faults, diverse fault sources and unclear fault causes [10]. Usually other methods are combined with its analysis methods to achieve the fault diagnosis of the hydraulic system. In this paper, the fault tree is combined with machine learning. Machine learning can be widely used in various fields because of its self-learning ability, which can effectively automatically analyze large amounts of data and obtain feature samples [11]. Therefore, the combination of technology and hydraulic control detection with fault diagnosis can greatly improve the error recognition rate. The literature shows that machine learning aided by data can have the best impact on the learning of the task at hand [12]. However, sometimes the use of auxiliary data can reduce learning efficiency, a problem known as negative transfer. Machine learning can be made more stable by utilizing multi-tasking resources during learning to avoid negative transfer [13]. This method mainly adjusts the weight of hydraulic control detection and fault diagnosis of construction machinery to narrow the difference between the target domain and the source domain, and finds the most suitable model for the target scope by comparing the weighted data. The basic idea is that adding weights to the training sample in the source domain will have a significant impact on the objectives of the learning task [14].
3.1 Decision tree algorithm
The decision tree consists of decision nodes, branches and leaf nodes. The highest node in the decision tree is called the root, and the root is the beginning of each decision tree node. The child node on each node in the decision tree is related to the algorithm decision tree. ID3 algorithm uses information in the decision tree of each node to obtain the feature selection criteria and repeatedly constructs the decision tree, which is a top-down, divide-and-rule inference process. ID3 algorithm is improved to obtain C4.5 algorithm, which optimizes the selection with the highest information acquisition rate of attribute test. It can not only deal with different attributes, but also deal with the continuity of attributes. Algorithmic tools can also be used to work with highly skewed or polymorphic digital data, as well as sequential or unarranged data. The algorithm uses hyperplanes to divide the space into two parts. Each part divides the current space into two parts, dividing repeatedly until the requirement is met, and finally each node divides the regions in the space.
The regression tree model can be expressed as follows:
Assuming that CART regression tree can be divided into M units corresponding to spatial input, and each unit has a fixed output value, the regression tree model can be expressed as follows:
The optimal value for each element is the partition of all vector inputs, and the corresponding representation output value, i.e
$$\widehat{{c}_{m}}=avg\left({y}_{i}\right|{\text{x}}_{\text{i}}\in {\text{R}}_{\text{m}})$$
1
In the input space division, input the optimal variable j into Eq. (1) to obtain the optimal segmentation point s.
$$\widehat{{c}_{1}}=avg\left({y}_{i}\right|{x}_{i}\in {R}_{1}(j,s))$$
2
$$\widehat{{c}_{2}}=avg\left({y}_{i}\right|{x}_{i}\in {R}_{2}(j,s))$$
3
3.2 BP neural network algorithm
The basic algorithm of BP includes two operations: forward release and error reverse transmission. In other words, the error of the output is calculated in the opposite direction from the output to the input, and the output is adjusted by adjusting the weight and threshold, and the output is input in the reverse direction. When published forward, the input signal is switched nonlinear to generate the output signal. Error reverse sending returns the input layer through the hidden layer and divides the error into cells in each layer. By adjusting the intensity of communication between input and output, hiding nodes, you can reduce errors in the direction of descent. Then the neural network can be trained to deal with the information transmitted in nonlinear so that the error of similar sample input is minimized.
The output of the node displayed in the hidden layer is shown in Formula (4), and formula (5) is the output of the node hidden layer displayed.
$${y}_{j}=f\left({net}_{j}\right)$$
4
$${O}_{k}=f\left({net}_{k}\right)$$
5
You can adjust the weight based on the error value to get the correct output if the desired output is not available through forward deployment. Suppose the sample set of BP neural network training input group:
$${E}^{\left(p\right)}=\frac{1}{2}\sum _{k=1}^{m}{({d}_{k}^{p}-{y}_{k}^{p})}^{2}$$
6
Input all the data into the neural network model, and the total error is:
$${E}_{s}=\sum _{p=1}^{N}{E}^{p}$$
7
The minimum average error box is the learning standard, and the ultimate goal of learning is to ensure that the error between the actual output and the expected output is less than the error tolerance through continuous updating.
$${w}_{jk}^{N+1}={w}_{jk}^{N}+\eta {\sigma }_{o}{y}_{j}$$
8
The weight of the output layer is corrected to:
$${w}_{jk}^{N+1}={w}_{jk}^{N}+\eta {\sigma }_{h}x$$
9
Given an infinitely differentiable activation function g(·) in any interval and an arbitrarily small error ε(ε < 0), in the case of any assignment:
$$‖{H}^{T}B-T‖=\epsilon$$
10
Formula (13) is solved to obtain
$$\widehat{B}={\left(H{H}^{T}\right)}^{-1}HT$$
11
Based on ridge regression principle, error can be eliminated by input correction coefficient λ, from which output weight can be obtained:
$$\widehat{B}={(H{H}^{T}+\lambda I)}^{-1}HT$$
12
So, the output function of the extreme learning machine is
$$\text{y}（\text{x}）=\text{h}\left(\text{x}\right)\widehat{\text{B}}$$
13
Duplicate nodes of hidden layer may appear in the immediate occurrence of duplicate nodes of hidden layer. Therefore, in order to make the model more accurate, ELM needs a large number of hidden nodes.
3.3 Parameter Learning
The learning structure of Bayesian networks can be defined as an optimization model: where G is the set of all variables in the sample D data set which may be related to each other; Ω is a set of constraints that must be met in a network structure. F is a scoring function to evaluate the network structure. Generally speaking, there are three types of structural learning methods, which are based on score search function, constraint and random sampling.
According to the uniform distribution, the K2 score of formula (14) can be obtained.
$${F}_{k2}(G/D)=logP\left(G\right)+{\sum }_{i=1}^{n}{\sum }_{j=1}^{{q}_{i}}\left[log\right(({r}_{i}-1)!/({m}_{ij}+{r}_{i}-1)!)+{\sum }_{k=1}^{{r}_{i}}log({m}_{ijk}!)]$$
14
The likelihood conditional probability is shown in (15) :
$${\theta }_{ijk}={m}_{ijk}/{m}_{ij}$$
15
Parameter learning is a process to determine the relationship between random variables. Bayes structure G network and sample D data set are required to determine the conditional probability distribution in each node of the model. Maximum likelihood estimation and Bayesian estimation are the main methods.
The method of parameter learning in this paper adopts maximum likelihood estimation, and likelihood degree can be expressed as follows according to the relationship between parameters and data sets:
$$\text{L}({\theta }/\text{D},\text{G})=\text{P}(\text{D}/{\theta },\text{G})$$
16
Maximum likelihood estimation selects the parameter that maximizes the likelihood function value, namely:
$${\theta }^{*}=argmaxL(\theta /D,G)$$
17
According to the independent co-distribution hypothesis of the data set and the structural characteristics of the Bayesian network, it can be obtained:
$$\text{L}({\theta }/\text{D},\text{G})=\text{l}\text{o}\text{g}{\prod }_{\text{i}=1}^{\text{m}}\text{P}({\text{d}}_{\text{i}}/{\theta })$$
18
3.4 Simulation results and evaluation
The characteristic curves of electric and hydraulic valve openings are normalized as input to the DA-ELM model. The activation function and node number model which affect the performance of DA-ELM are analyzed experimentally. See Fig. 1.
The training learning of XGBoost fault diagnosis model is obtained through the training set data. After learning and training, the test set is input into the "fault model application" module for model testing. The accuracy rate, recall rate, F1 measure and confusion matrix are used for quantitative estimation in the "model test evaluation" module. The evaluation confusion matrix of the XGBoost fault diagnosis model is shown in Table 1.
| Type | True situation | Normal | Slight | Moderate | Serious | Mean value |
|---|---|---|---|---|---|---|
| Forecast result | Label | 0 | 1 | 2 | 3 | |
| Normal | 0 | 75 | 5 | 4 | 3 | |
| Slight | 1 | 6 | 66 | 4 | 1 | |
| Moderate | 2 | 7 | 7 | 89 | 7 | |
| Serious | 3 | 3 | 5 | 6 | 89 | |
| Accuracy rate (%) | Precision | 87.61 | 87.47 | 82.36 | 87.75 | 86.29 |
| Recall rate (%) | Recall | 85.67 | 84.55 | 86.78 | 86.79 | 85.95 |
| F1 measure (%) | F1-score | 86.63 | 85.99 | 83.98 | 87.26 | 85.97 |
Based on the original software, changes in the accuracy of motor faults are observed by adjusting the size of the convolutional layer and the number of channels in the convolutional layer. The relationship between the accuracy of motor faults and the size of the convolutional kernel under different channels is shown in Table 2:
Table 2 Relationship between motor fault accuracy and convolution kernel size in different channels (%)
According to the results in Table 2, when the number of channels in the package is in a certain layer, the error rate will become higher as the number of convolution nuclei goes from minimum to maximum. The convolution kernel dimension is very small and completely covers the data input, which leads to a significant increase in the convolution process, which is a long process model and leads to the degradation of network performance. When the dimension of convolution kernel exceeds the normal range, the features of input data cannot be accurately identified, thus the extraction function may be unclear and repetitive, resulting in the phenomenon of overfitting.
4.1 Overall architecture design of monitoring system
Machine sampling working environment is poor, the working conditions are more complex, powerful technology, high reliability, we must well analyze the actual working conditions, make the system more practical, safe and efficient. Based on the principle of comprehensive, step-by-step control, the system is based on a top-down design concept, gradually implemented to each unit, making the system more adaptable.
Comprehensive knowledge of hardware and software system structure to ensure easy maintenance and flexible expansion. The control system should maintain the necessary reserved i/o ports, and select a controller with some additional capacity to ensure that the terminal will automatically control the system function in the future and ensure program upgrading.
The overall software program composition and specific functions required are shown in Fig. 2.
Figure 2 introduces the five main functions of the test software of the servo valve testing platform, including system configuration, data simulation and acquisition, characteristic analysis of the electro-hydraulic servo valve, data storage and fault diagnosis.
4.2 Fault Diagnosis Module Design
Hydraulic components manufacturing and installation accuracy is high, so that the hydraulic components internal damage state, and eventually lead to performance changes.
The hydraulic system is gradually losing in operation. Major system failures can result from wear and tear due to frequent movement of the engine's hydraulic cylinders, or changes in the operating system, such as clogging due to pollution from mixing hydraulic fluids under complex operating conditions.
A sudden accident. Hydraulic components can be affected by abnormal conditions such as external forces. For example, an engine colliding with solid rock under harsh working conditions can cause a sudden stop that cannot be effectively estimated based on the model Settings.
The implementation steps are shown as follows.
It is assumed that a total of n groups of data are collected, and each group contains p features. The feature data is written in the form of nxp dimension matrix, as shown in Eq. (19).
$$\text{X}=({\text{X}}_{1},{\text{X}}_{2},...,{\text{X}}_{\text{p}})$$
19
Eliminating dimensionality with raw data is the preparatory stage of processing raw data to avoid significant impact on the data processing process due to a large number of valuable data. Data standardization can be expressed as follows:
$${x}_{ij}^{*}=\frac{{x}_{ij}-\overline{{x}_{j}}}{\sqrt{var\left({x}_{j}\right)}}$$
20
The average value is:
$$\overline{{x}_{j}}=\frac{1}{n}\sum _{i=1}^{n}{x}_{ij}$$
21
In the operation of a hydraulic system, it is necessary to obtain different main output curves over time during the operation of the system, which can be used for further prediction. By simulating the hydraulic test platform and gradually adjusting the simulation from small to large unit leakage test bench, the seal hydraulic cylinder leakage is simulated. With the increase of time, the leakage area keeps increasing, and the main output characteristics change with time, as shown in Fig. 3.
According to the relationship between the internal leakage parameters and the main characteristics of the model of the internal leakage parameters, the change curve can be obtained from the parameter leakage in the hydraulic cylinder.
4.3 System Fault Diagnosis Performance and Effect
In essence, SVM classification model is the pattern recognition and classification recommendation of data operation errors. In this paper, the tool libsvm-3.23 is used to build a classification model in MATLAB. Training sample 2 has a total of 300 data, including 70 jack failures, 60 column failures, 40 acting valves and 80 emulsion pump failures. Table 3 shows the distribution of training sample 2.
| Fault type | Jack failure | Column failure | Operating valve failure | Emulsion pump failure | Total |
|---|---|---|---|---|---|
| Training set | 40 | 25 | 30 | 35 | 125 |
| Test set | 20 | 30 | 30 | 40 | 125 |
| Total | 60 | 55 | 60 | 75 | 250 |
When selecting kernel function, linear kernel function is difficult to process high-dimensional data, multinomial kernel function and Sigmoid kernel function set parameter function is very complicated. A good classification model requires the selection of appropriate parameters c and g. In this section, when selecting parameters, the weights of factor c and kernel parameter g are first evaluated empirically, and then adjacent values are selected to combine the parameters. By combining parameters of 4×4 = 16 types, faults are classified and constructed. c and g with the highest accuracy are selected as parameters for model construction, as shown in Table 4.
| c g | 0.5 | 1.0 | 1.5 | 2.0 |
|---|---|---|---|---|
| 0.15 | 97.7% | 96.9% | 94.5% | 98.5% |
| 0.20 | 73.7% | 98.5% | 99.2% | 97.7% |
| 0.25 | 96.1% | 98.5% | 100.0% | 97.7% |
| 0.30 | 98.5% | 93.7% | 96.1% | 96.9% |
As can be seen from Table 4, in various parameter groups, when the highest classification accuracy is 100%, the corresponding values c and g are 1.5 and 0.25, respectively. Although the choice of c and g methods is very subjective, it guarantees 100% accuracy, so this paper chooses 1.5 times the value of penalty coefficient c in the function and 0.20 times the value of the kernel parameter g.
Three sets of data trained by algorithms that have the highest success rate in diagnosing one type of fault and the lowest mixing error rate. Especially for the hydraulic cylinder with wear, the success rate of fault diagnosis is low, because the influence of wear press on the hydraulic system output is more complex. Table 5 Success rate of fault diagnosis algorithm.
| Algorithm data feature set | Failure data of hydraulic test platform | Machine learning mimics simulation data | Traditional model simulation data |
|---|---|---|---|
| Comprehensive fault diagnosis accuracy | 96.6% | 82% | 81% |
Modeling was carried out on LS simulation step signal hydraulic braking system of mine hoist under constant deceleration braking mode. The sampling period of the system was set as 0.01s as the sampling period of the model, and the simulation time was 30s. A unit was added in Os, and the simulation results were shown in Fig. 4:
Under constant deceleration braking mode, the proportional relief valve control response curve of LS mine hoist hydraulic braking system is shown in Fig. 4. As can be seen from the figure, when the steady-state error is 30%, the overshot amount of the system is 35.7%.
In this paper, taking the hydraulic pressure of construction machinery as the research object, BP neural network, MLP model of single hidden layer and multi-hidden layer, convolutional neural network and other methods are used to collect hydraulic data samples of construction machinery and fault diagnosis. The hydraulic fault diagnosis of construction machinery is studied, and the establishment of different models, numerical extraction, sample data and training diagnosis results are analyzed in detail. Firstly, the BP neural network algorithm in the field of machine learning is studied. Then, the fault classification results are inputted into the BP neural network model to train and detect errors and identify errors according to the classification error results. Then, the MLP(multi-hidden layer sensor model) model is studied, and a hydraulic fault diagnosis method based on MLP model is proposed. According to the detection and diagnosis results, the function of this method in hydraulic system fault diagnosis is verified. Finally, the neural network torsion (CNN) in the field of learning machine is studied and successfully applied to the hydraulic control detection and fault diagnosis method of construction machinery.
Compliance with Ethical Standards
Conflict of interest
The authors declare that they have no conflict of interests
Ethical approval
This article does not contain any studies with human participants performed by any of the authors.
Data Availability
Data will be made available on request.
- T. Wang, & Q. Wang, “An energy-saving pressure-compensated hydraulic system with electrical approach,” IEEE/ASME transactions on mechatronics, vol. 19, no. 2, pp. 570-578, 2013.
- Y. Wang, Q. Li, M. Chang, et al., “Research on fault diagnosis expert system based on the neural network and the fault tree technology,” Procedia Engineering, vol. 31, pp. 1206-1210, 2012.
- M. I. Jordan, & T. M. Mitchell, “Machine learning: Trends, perspectives, and prospects,” Science, vol. 349, no. 6245, pp. 255-260, 2015.
- M. Aghbashlo, W. Peng, M. Tabatabaei, et al., “Machine learning technology in biodiesel research: A review,” Progress in Energy and Combustion Science, vol. 85, p. 100904, 2021.
- Y. Lei, F. Jia, J. Lin, et al., “An intelligent fault diagnosis method using unsupervised feature learning towards mechanical big data,” IEEE Transactions on Industrial Electronics, vol. 63, no. 5, pp. 3137-3147, 2016.
- U. R. Acharya, S. L. Oh, Y. Hagiwara, et al., “A deep convolutional neural network model to classify heartbeats,” Computers in biology and medicine, vol. 89, pp. 389-396, 2017.
- C. W. Coley, W. Jin, L. Rogers, et al., “A graph-convolutional neural network model for the prediction of chemical reactivity,” Chemical science, vol. 10, no. 2, pp. 370-377, 2019.
- K. A. Bhavsar, A. Abugabah, J. Singla, et al., “A comprehensive review on medical diagnosis using machine learning,” Computers, Materials and Continua, vol. 67, no. 2, p. 1997, 2021.
- J. Dai, J. Tang, S. Huang, & Y. Wang, “Signal-based intelligent hydraulic fault diagnosis methods: Review and prospects,” Chinese Journal of Mechanical Engineering, vol. 32, no. 1, pp. 1-22, 2019.
- E. Ruijters, & M. Stoelinga, “Fault tree analysis: A survey of the state-of-the-art in modeling, analysis and tools,” Computer science review, vol. 15, pp. 29-62, 2015.
- Z. H. Zhou, “Learnware: on the future of machine learning,” Frontiers Comput. Sci., vol. 10, no. 4, pp. 589-590, 2016.
- K. T. Butler, D. W. Davies, H. Cartwright, et al., “Machine learning for molecular and materials science,” Nature, vol. 559, no. 7715, pp. 547-555, 2018.
- S. Sun, Z. Cao, H. Zhu, & J. Zhao, “A survey of optimization methods from a machine learning perspective,” IEEE transactions on cybernetics, vol. 50, no. 8, pp. 3668-3681, 2019.
- V. Popovici, W. Chen, B. D. Gallas, et al., “Effect of training-sample size and classification difficulty on the accuracy of genomic predictors,” Breast Cancer Research, vol. 12, no. 1, pp. 1-13, 2010.