**5. Influencing factors of support scheme design**

#### **5.1 Evaluation methods for neural network weights**

The choice of a system anchor support scheme in an underground plant is influenced by several factors, but it is not yet clear which is the main factor. Based on the intelligent design model proposed above, the different degrees of influence of the plant span and strength-stress ratio on the selection of the system anchor support scheme can be further explored by analyzing the weights between the neurons. The relationship can be described with the help of the following indicators [38].

1.Correlation significance coefficient *rij*:

$$r\_{ij} = \sum\_{k=1}^{p} \frac{\alpha\_{ki}(\mathbf{1} - e^{-\mathbf{x}})}{\mathbf{1} + e^{-\mathbf{x}}} \tag{21}$$
  $\mathbf{x} = \alpha\_{jk}$ 

where *i* is the input unit of the neural network, *i* = 1, … … , *m*; *j* is the output unit of the neural network, *j* = 1, … … , *n*; *k* is the hidden unit, *k* = 1, … , *p*; *ωki* is the weight coefficient between neuron *i* in the input layer and neuron *k* in the hidden layer, and *ωjk* is the weight coefficient between neuron *j* in the output layer and neuron *k* in the hidden layer.

2.Correlation index *Rij*:

$$\begin{aligned} R\_{\vec{\imath}\vec{\jmath}} &= \left| \frac{\mathbf{1} - e^{-\mathbf{y}}}{\mathbf{1} + e^{-\mathbf{y}}} \right| \\ \mathbf{y} &= r\_{\vec{\imath}\vec{\jmath}} \end{aligned} \tag{22}$$

3.Absolute impact factor *Sij*:

$$S\_{\vec{ij}} = \frac{R\_{\vec{ij}}}{\sum\_{i=1}^{m} R\_{\vec{ij}}} \tag{23}$$

The absolute influence coefficient *Sij* can be used to evaluate the influence of different input units on the output result, with a higher value of *Sij* for an input unit indicating a greater influence on the result.


#### **Table 10.**

*Neural network weights of the intelligent design model.*


#### **Table 11.**

*Weight of the support scheme for the ratio of span and strength stress.*

#### **5.2 Discussion of support scheme impact factors**

The weights of the neural network model in Section 4 are shown in **Table 10**.

The influence weights of the plant span and strength-stress ratio on the 6 system anchor support schemes are shown in **Table 11**.

As seen from **Table 11**, the weights of the strength-stress ratio on the results in the intelligent support design model are greater than the weights of the plant span. Therefore, when considering only the plant span and the strength-stress ratio, the variation in the strength-stress ratio has a greater influence on the choice of the anchor support scheme for the underground plant.
