Multivariate Nonlinear Regression Model Research Articles

Adjuvant surgical decision-making system for lumbar intervertebral disc herniation after percutaneous endoscopic lumber discectomy: a retrospective nonlinear multiple logistic regression prediction model based on a large sample

Lumbar disc herniation (LDH) is a common condition that can affects an individual' quality of life. In patients for whom conservative treatment is ineffective after 3 months, surgical treatment, such as percutaneous endoscopic lumbar discectomy (PELD), is recommended. Because PELD is minimally invasive and produces thorough nerve root decompression, both surgeons and patients often prefer it to other techniques. Surgeons find it challenging to prevent postoperative recurrent LDH (rLDH) when they use PELD. We created and verified a model for evaluating patients' recurrence risk factors before surgery so that surgeons can choose other surgical techniques when necessary. Retrospective study. One thousand eight hundred seven patients who underwent PELD at our hospital between 2012 and 2015 were enrolled. The main outcome measure was rLDH at any follow-up time point. Data were retrospectively analyzed for 1807 patients who underwent PELD at our hospital at some point between 2012 and 2015; all patients had been monitored for at least 5 years after surgery. They were divided into a recurrence group and a nonrecurrence group. Clinical and radiological risk factors were assessed over time to determine their correlations with recurrence and to exclude less important factors. A nonlinear multivariate logistic regression model was established to predict the recurrence rate before surgery. A total of 1706 patients were monitored after PELD; data were missing for 101 additional patients. The total recurrence rate was 10.38%, and the most common time from surgery to recurrence was 1 year. Ten risk factors were assessed and included in the analysis. Regarding clinical risk factors, patients with hypertension (p < .001; correlation coefficient R [R]=0.235; odds ratio [OR]=4.749), diabetes (p < .001; R=0.381; OR=16.797), a history of smoking (p < .001; R=0.347; OR=9.012), and a history of performing intense physical labor (p < .001; R=0.409; OR=19.592) had a higher recurrence rate. Regarding radiological risk factors, patients with disc degeneration (Pfirrmann grade III) (p < .001; R=0.228; OR=4.919), Modic changes (level 2) (p < .001; R=0.309; OR=7.934), herniation in the form of extrusion (p < .001; R=0.365; OR=12.228), a higher disc height index (DHI) (p < .001; R=0.336), and a larger segmental range of motion (p < .001; R=0.243) had a higher recurrence rate. When the lumbar motion angle was negative (p < .001; R=0.318; OR=13.680), the recurrence rate was high. The overall accuracy of the final model was 97.6% (1665 of 1706). The recognition rate for non-rLDH cases was 99.0% (1514 of 1529), and the rate for rLDH cases was 85.3% (151 of 177); the AUC was 0.9315. A simple model was used. For those patients with postoperative trauma (p < .001; R=0.382; OR=13.680), a comparison model was established, and the corresponding recurrence rate was 23.0% ± 25.0% (0-76%). A large cohort of patients underwent long-term monitoring, and 11 risk factors were verified for assessing each patient's risks before surgery to predict the postoperative recurrence of LDH following PELD. The risk of recurrence may be effectively reduced with the use of alternative surgical techniques in high risk cases.

Multiple nonlinear regression model of cutting force for C/SiC composites by laser‐assisted micromachining

AbstractBased on the response surface methodology (RSM), laser‐assisted micromachining (LAMM) experiments of carbon fiber reinforced silicon carbide matrix(C/SiC) composites have been carried out to obtain the data of three‐direction cutting forces Fx, Fy, Fz, and result‐force F in the cutting process. A stepwise regression method is used to screen out the factors that significantly affected the cutting force, and hence a multivariate nonlinear mathematical regression model has been established. Through the analysis of variance and regression analysis, the R2 of the regression model is the cutting force Fx: 0.975, the cutting force Fy: 0.970, the cutting force Fz: 0.972, the resultant force F: 0.987, and the probability p &gt; F‐values are all less than 0.001. The model analysis results show that the main effects and interactions of the three factors of laser power, cutting speed, and cutting depth significantly influence the cutting force. The main effect of the most considerable influence on the resultant F is cutting depth, followed by cutting speed and laser power, and the interactive effect is cutting speed and cutting depth. The experimental results show that the model prediction results are in good agreement with the experimental results.

Agro waste as a source of bioactive glass for targeted drug delivery and bone implantation

The present work describes the synthesis of biocompatibe alkali-free glasses utilizing the agricultural waste rice husk and eggshells. A series of glasses with basic composition SiO2–P2O5–MgO–CaO has been synthesized using melt-quench technique, where P2O5 is partially replaced by MnO. These glasses are characterized to assess their in-vitro reactivity with the help of X-Ray diffraction, scanning electron microscopy, energy dispersive X-ray spectroscopy, fourier transform infrared spectroscopy, and microwave plasma atomic emission spectroscopy techniques before and after soaking in the simulated body fluid solution. Further, drug loading and release studies are carried out with the help of UV–Visible spectrophotometer. Wherefore, a multivariable non-linear regression model has been developed to predict the release rate at different manganese oxide concentrations. The proposed model shows a high Pearson correlation coefficient between the observed and predicted absorbance. These results designate that low-cost agro-waste derived alkali-free glasses have potential in medical applications in drug delivery as well as it helps in environmental pollution management.

Long-term evaluation and calibration of three types of low-cost PM2.5 sensors at different air quality monitoring stations

To evaluate the performance of low-cost PM2.5 sensors and develop calibration models for correcting for the PM2.5 sensor data (PM2.5,S), field comparison tests were conducted based on Met One BAM-1020 data at various locations using long-term (≥one year) data of Plantower PMS5003, Sensirion SPS30, and Honeywell HPMA115S0 PM2.5 sensors. Both multivariate linear regression (MLR) and non-linear regression (NLR) models using hourly RHs and original sensor PM2.5 data as parameters were able to obtain accurate calibrated hourly PM2.5 values with MNBs (mean normalized biases) less than about ±10% and MNEs (mean normalized errors) less than about 30% for all three types of PM2.5 sensors at all monitoring locations. On the other hand, the MNB and MNE of the calibrated 24-hr average PM2.5 data for the two models were less than ±13% and 20%, respectively. Moreover, the slope, intercept, and R2 of the linear regression line of the calibrated 24-hr average PM2.5 and BAM-1020 data were as good as 1.0 ± 0.1, 0.0 ± 2.0 μg/m3, and ≥0.88, respectively. Therefore, these well-calibrated sensors can well be served for education and information (MNE<50%), hotspot identification and characterization (MNE<30%), and personal exposure study (MNE<30%) purposes, and even supplement the existing daily PM2.5 data of the air quality monitoring stations (MNE<20%).

Multivariate Quadratic Nonlinear Regression Model of the Ultimate Pull-Out Load of Electrohydraulic Expansion Joints Based on Response Surface Methodology

Electrohydraulic expansion joining has great potential for joining the light weight and high strength thin-walled pipes due to its high strain rate. Based on the central composite design (CCD) of response surface methodology, multiple experiments of electrohydraulic expansion joining process were performed. The multivariate quadratic nonlinear regression model between process parameters (discharge voltage, wire length, and wire diameter) and the ultimate pull-out load of the joints was established. The results revealed that discharge voltage, wire length and wire diameter all had a significant effect on the ultimate pull-out load. The discharge voltage had the most significant effect. The interaction between the discharge voltage and the wire diameter had a significant effect on the ultimate pull-out load. The optimal parameter combination (discharge voltage = 6 kV, wire length = 10 mm, wire diameter = 0.833 mm) was obtained and verified through the experiments. This study would provide guidance for the choice of the process parameters in real applications.

A Novel Load Forecasting Model for Automatic Fault Clearance in Secondary Distribution Electric Power Grid using an Extended-Multivariate Nonlinear Regression

Smart grid is an emerging platform adopted by many electric power utility companies to enhance proper service delivery as well as cost-effective operations. Automatic Fault Detection and Clearance (AFDC) is a part of intelligent technology initiatives established on Tanzania’s grid aiming at detecting, managing, and handling fault with little or without human intervention. Being one of the components of AFDC, the Load Forecasting agency plays important role in feeding restoration and distributed energy resource agencies. Forecasting load demand profile in the AFDC requires a robust mechanism for accommodating both fault-based and random-walk situations. Unfortunately, existing LF mechanisms do not address these two issues robustly. Therefore, this research work aims to propose an AFDC-based LF (LF-AFDC) model whose time series data is characterized by random-walk distribution. Firstly, we establish design requirements using focus group discussion and literature review. Secondly, iterative validation is conducted to assess the suitability of the proposed LF mechanism in the intended context. Thirdly, the core forecasting part of the mechanism is achieved using the extended Multivariate Nonlinear Regression (e-MNLR) model. Findings indicate the capability of the proposed model to forecast the next load profile from the given fault-date, fault-time, restoration duration, Gross-Domestic Product (GDP), number of customers, and population information. Furthermore, the designed e-MNLR seems to outperform the ANN, SVM, LSTM, and MNLR models.

Modified artificial neural network model with an explicit expression to describe flow behavior and processing maps of Ti2AlNb-based superalloy

The elevated-temperature deformation behavior of Ti2AlNb superalloy was observed by isothermal compression experiments in a wide range of temperatures (950–1200 °C) and strain rates (0.001–10 $${\mathrm{s}}^{-1}$$ ). The flow behavior is nonlinear, strongly coupled, and multivariable. The constitutive models, namely the double multivariate nonlinear regression model, artificial neural network model, and modified artificial neural network model with an explicit expression, were applied to describe the Ti2AlNb superalloy plastic deformation behavior. The comparative predictability of those constitutive models was further evaluated by considering the correlation coefficient and average absolute relative error. The comparative results show that the modified artificial network model can describe the flow stress of Ti2AlNb superalloy more accurately than the other developed constitutive models. The explicit expression obtained from the modified artificial neural network model can be directly used for finite element simulation. The modified artificial neural network model solves the problems that the double multivariate nonlinear regression model cannot describe the nonlinear, strongly coupled, and multivariable flow behavior of Ti2AlNb superalloy accurately, and the artificial neural network model cannot be embedded into the finite element software directly. However, the modified artificial neural network model is mainly dependent on the quantity of high-quality experimental data and characteristic variables, and the modified artificial neural network model has not physical meanings. Besides, the processing maps were applied to obtain the optimum processing parameters.

A multivariate nonlinear regression model for the resistance power of a light rail vehicle

In Turkey, 20% of energy use caused by transportation. Light rail transportation has developing nowadays. However, it is very important to issue having information about the energy consumption of the light rail vehicle according to different both vehicle and railway situations. &#x0D; In Turkey, approximately 20% of the energy expended is spent on transportation. Light rail transportation technology is still in an evolving process. In this development process, it is crucial to have information about the energy consumption of the light rail vehicle according to the different situations of both the vehicle and the railway.&#x0D; It is necessary to predict the power losses that will occur under different driving conditions sensitively to ensure energy efficiency in light rail systems. The most important of these power losses is the resistance loss caused by contact with the route. Resistance loss is dependent multiple environmental conditions. The most important of these conditions can be listed as the weight of the light rail vehicle, the instantaneous speed of the vehicle, the curve of the route, the ramp slope of the route and the friction force arising from these conditions. Resistance loss is proportional and linearly dependent to some of these variables while others show reverse or nonlinear dependence. Due to these different types of dependencies, it is seen that a single multivariate nonlinear model is needed to explain the loss of resistance in all different conditions. In this study, a new and accurate model for resistance losses has been developed by fitting numerical values obtained from different scenarios to multivariate nonlinear regression model.

Experimental study on heat transfer in soil during heat storage and release processes

The soil temperature distribution characteristics were experimentally studied, during heat storage and release processes in the fine-grained sand with different moisture content. Results showed that, the soil temperature distributions in the process of soil heat storage and release are almost symmetric under the same excess temperature. The excess temperature is the main factor affecting the soil temperature change, and the influence of the moisture content variation is almost negligible. At each set of excess temperature, the soil temperature change in heat storage and release processes have a good logarithmic relationship with the heat transfer time, radial distance and longitudinal distance. The soil temperature change shows a linear increase trend with the excess temperature, and the soil temperature disturbance in radial direction is far greater than that in longitudinal direction. Based on the hypothetical multivariate nonlinear power function regression model, the experience formula for predicting the soil temperature was proposed. The prediction deviation was within 10 %.

Research on multivariate nonlinear regression model of specific energy of rock with laser drilling based on response surface methodology

Based on Box–Behnken Design of response surface methodology, multiple experiments of laser drilling (four centimeters thickness of sandstone) were performed by continuous fiber laser (Model MFSC) with a maximum laser power 1 kW to obtain the specific energy of rock with three experimental parameters of laser power (400∼800 W), irradiation time (6∼10 s) and spot diameter (0.01∼0.03 m). According to experimental results, a multivariate nonlinear mathematical regression model correlating the experimental parameters with the specific energy of rock was established and the accuracy of the model was found to be good by variance analysis. Univariate analysis of the model revealed that laser power had the greatest influence on the specific energy of rock, while irradiation time affected the least. Considering the interaction effect among three experimental parameters, calculation results of the model exhibited that the combination of laser power and irradiation time, as well as that of laser power and spot diameter, had obvious influence on the specific energy of rock, while the combination of irradiation time and spot diameter showed little effect. After that, the optimal parameter combination (P=768.9 W, t=8.3 s, d=0.01 m) was obtained. Five repeated experiments under the optimal parameter combination clearly illustrated good agreement between predicted values and measured values. This paper proposed a multivariate nonlinear model based on the response surface method with high precision in predicting the specific energy of rock, which could help reduce the specific energy of rock and improve the efficiency of rock breaking.

Cloning data with unchanged estimates of estimable non-linear functions of parameters

Non-linear regression models occur in the fields of biology, banking, economics, and sociology, population and biological growth. The absolute growth, growth of humans, and most importantly, an economic variable is appropriately described by non-linear regression models. In this article, we present cloned datasets for bivariate and multivariate non-linear regression models with the same non-linear regression fit. The application of such cloned datasets is used for the confidentiality of sensitive data for publication purposes. In this article, we present cloned data sets which will yield the same fitted non-linear regression models.

Characterization of electrical current and liquid droplets deposition area in a capillary electrospray

A comprehensive study was conducted to investigate the effects of applied voltage, distance between needle tip and counter electrode, liquid flow rate, pH and conductivity of water on the electrical current in a capillary electro-spray. Atomic Force Microscopy (AFM) was employed to measure the size distribution of water droplets, which presented a mean diameter of 299 ​± ​76 ​nm. Statistical analysis using three-way ANOVA and nonlinear multivariate regression model were employed to study the interactions of applied voltage, distance between needle tip and counter electrode, and liquid feed rate with electrical current, and to develop regression equations to predict the current. It was found that electrical current was significantly (p ​< ​0.05) influenced by applied voltage (0 to ±10 ​KV) and the distance (2, 3 and 4 ​cm), but not by the liquid feed rate (1–10 ​μl/min) within the range investigated in this work. A water contact indicator was employed for the first time in the electrospray system to investigate the effect of parameters mentioned above on the liquid droplets deposition area. The experimental data demonstrated that increasing the distance of the capillary needle, liquid feed rate, pH, and conductivity, resulted in increased electrosprayed area.

A study on the interdependence of heavy metals while contributing to groundwater pollution index.

The contribution of heavy metal to the groundwater pollution index (m-HPI) is dependent on other heavy metals present in it. This contribution may be synergistic or anti-synergistic (antagonistic) depending upon the constituent matrix. Both heavy metal type and its concentration are important. m-HPI, a variant of heavy metal pollution index in water, may be calibrated against USEPA hazard index (HI) using a generic multivariate non-linear regression (MVNLR) model. Excellent correlation may be obtained between HI and m-HPI through optimization of normalized weightage factors of constituent metals that contribute to m-HPI. MVNLR model was employed on groundwater samples of ten sites having different heavy metal matrix. The synergistic/antagonistic contribution of heavy metals to m-HPI was well discernible at each site. This study clearly showed that the individual contribution of a particular heavy metal to pollution index might be altered (enhanced or reduced) in the presence of other heavy metals. A calibrated MVNLR model was successfully used for predicting the hazard index (HI) of water samples.

A food system optimization model Analysis

The optimization of food system is studied in this paper. Today's food system, while highly efficient and profitable, is generally unstable, and this instability is reflected in equity and sustainability. In our understanding, to achieve fairness, it is necessary to eliminate monopoly, distribution according to demand, production according to demand and safe reserve; Sustainability is mainly for the protection of water resources and natural environment resources. In order to evaluate the security of the food system, we set up the food system security evaluation system. First, identify the internal factors that affect the security of the food system, that is, whether the yield and nutritional structure of agricultural and livestock products in a given country can meet the needs of its people. Establish, a multivariate nonlinear regression model according to the linear relationship between various nutrients and food security. Then, the external factors affecting the security of the food system and the influences of natural and social environment on the food system were identified. The index system is established, namely, the safety index sequence of each indicator is obtained, and the influence degree of external factors is given. Finally, combining with the weight ratio of external and internal causes, the food system security evaluation system is obtained.

МОДЕЛЬ QSPR ДЛЯ ПРОГНОЗА ОКТАНОВЫХ ЧИСЕЛ УГЛЕВОДОРОДОВ РЯДА АЛКЕНОВ ПО ТОПОЛОГИЧЕСКИМ ХАРАКТЕРИСТИКАМ МОЛЕКУЛ

Relevance Modeling, optimization and design of processes in the technology of oil refining and organic synthesis, as well as the use of new chemical and technological processes involv ing hydrocarbon systems with the calculation of physical and chemical properties (РCS) of liquids and gases. The accuracy of calculations plays an observed role in the calculations of oil refining and organic synthesis processes, as well as approximate reactor processes, fractionation and heat exchange equipment. Despite the availability of РCS databases, the search for adequate initial data takes a long time at the stage of the entire development or design process, since the РCS of many hydrocarbons remains unknown. Aim of research To develop a quantitative Structure-Property Ratio (QSPR) model for octane hydrocarbons of a series of alkenes. Research methods To predict the octane numbers of normal and substituted alkenes — components of catalytic cracking, a nonlinear multivariate regression model Quantitative Structure-Property Relationship (QSPR) is proposed. The objects of the study were 30 hydrocarbons of a number of alkenes, selection in the basic and test samples, made randomly using computer data of physical and chemical properties. The model associates a set of descriptors with the octane numbers — the topological characteristics of their molecular graphs: the Wiener index, the Randich index and the magnitude of the quadratic dependence of bone structure, which affect the octane numbers and reflect the main structural and chemical factors, such as the length and branching of the carbon skeleton, and sensitive parameters. molecules. Results The adequacy of the models was confirmed by statistical data processing, so the coefficient of determining the models is 0.856. For the quality characteristics of the QSPR model, the multiple correlation coefficient r = 0.925 was calculated, which suggested a force relationship between the proposed topological characteristics of hydrocarbon molecules and their octane numbers. To assess the statistical stability of the model, a correlation correction was used. The maximum absolute and relative errors for octane number sampling tests are 4.0 units and 4.1 %, respectively. The statistical calculation, which makes it possible to judge the adequacy of the predicted indicators, their compliance with the reference data, is the standard regression error of 6.5 units. The small value of the standard error of the regression in comparison with the values of the dependence of the applicable adequacy of the proposed model. The model adequately uses the octane numbers of linear and branched alkenes and can be used to predict the octane numbers of gasoline components.

Modelling of River Flow Using Particle Swarm Optimized Cascade-Forward Neural Networks: A Case Study of Kelantan River in Malaysia

Water resources management in Malaysia has become a crucial issue of concern due to its role in the economic and social development of the country. Kelantan river (Sungai Kelantan) basin is one of the essential catchments as it has a history of flood events. Numerous studies have been conducted in river basin modelling for the prediction of flow and mitigation of flooding events as well as water resource management. This paper presents river flow modelling based on meteorological and weather data in the Sungai Kelantan region using a cascade-forward neural network trained with particle swarm optimization algorithm (CFNNPSO). The result is compared with those trained with the Levenberg–Marquardt (LM) and Bayesian Regularization (BR) algorithm. The outcome of this study indicates that there is a strong correlation between river flow and some meteorological and weather variables (weighted rainfall, average evaporation and temperatures). The correlation scores (R) obtained between the target variable (river flow) and the predictor variables were 0.739, −0.544, and −0.662 for weighted rainfall, evaporation, and temperature, respectively. Additionally, the developed nonlinear multivariable regression model using CFNNPSO produced acceptable prediction accuracy during model testing with the regression coefficient (R2), root mean square error (RMSE), and mean of percentage error (MPE) of 0.88, 191.1 cms and 0.09%, respectively. The reliable result and predictive performance of the model is useful for decision makers during water resource planning and river management. The constructed modelling procedure can be adopted for future applications.

Defining accurate delivery dates in make to order job-shops managed by workload control

Workload control (WLC) is a lean oriented system that reduces queues and waiting times, by imposing a cap to the workload released to the shop floor. Unfortunately, WLC performance does not systematically outperform that of push operating systems, with undersaturated utilizations levels and optimized dispatching rules. To address this issue, many scientific works made use of complex job-release mechanisms and sophisticated dispatching rules, but this makes WLC too complicated for industrial applications. So, in this study, we propose a complementary approach. At first, to reduce queuing time variability, we introduce a simple WLC system; next we integrate it with a predictive tool that, based on the system state, can accurately forecast the total time needed to manufacture and deliver a job. Due to the non-linearity among dependent and independent variables, forecasts are made using a multi-layer-perceptron; yet, to have a comparison, the effectiveness of both linear and non-linear multi regression model has been tested too. Anyhow, if due dates are endogenous (i.e. set by the manufacturer), they can be directly bound to this internal estimate. Conversely, if they are exogenous (i.e. set by the customer), this approach may not be enough to minimize the percentage of tardy jobs. So, we also propose a negotiation scheme, which can be used to extend exogenous due dates considered too tight, with respect to the internal estimate. This is the main contribution of the paper, as it makes the forecasting approach truly useful in many industrial applications. To test our approach, we simulated a 6-machines job-shop controlled with WLC and equipped with the proposed forecasting system. Obtained performances, namely WIP levels, percentage of tardy jobs and negotiated due dates, were compared with those of a set classical benchmark, and demonstrated the robustness and the quality of our approach, which ensures minimal delays.

Модель QSPR для прогноза динамической вязкости аренов по топологическим характеристикам молекул

Prediction of the dynamic viscosity of saturated hydrocarbons vapors is an important step in the calculation of various processes and apparatuses in chemical technology. In order to quickly determine the dynamic viscosity without resorting to the use of expensive equipment, methods of mathematical modeling are currently used. To predict the dynamic viscosity of saturated arenas vapors, a nonlinear multivariate regression model QSPR is proposed. The model associates with a dynamic viscosity a set of descriptors – the topological characteristics of molecular graphs: the Randic index, the Wiener index, and also the functions of the eigenvalues of the topological matrix of the molecule, which reflect the main structural and chemical factors, such as branching, length of the carbon skeleton and energy parameters of molecules, for example, Hückel’s perturbation spectrum of molecules, as well as affecting dynamic viscosity. The objects of research used arenas. The studied sample included 40 hydrocarbons of a number of arenas. The proposed model adequately describes the dynamic viscosity of saturated arenas vapors. The coefficient of determination of the model is 0.986. The average absolute and relative error for the test sample of HF is -2.46⋅10-7 cP and 1.83%, respectively. The model is applicable for engineering and scientific forecasts of the dynamic viscosity of various saturated arenas vapors.

Research on nonlinear prediction model of weld forming quality during hot-wire laser welding

In order to realize accurate prediction of the weld forming quality during hot-wire laser welding, in this study, a basic multivariate linear regression model regarding multiple process parameters and the weld forming quality is first established. And a residual analysis is then used to analyze the linear regression model. The results of the residual analysis illustrate that the influences of the hot-wire current, welding speed, and wire feeding rate on the weld forming quality are nonlinear, whereas the influences of the laser power and gap width on the weld forming quality are linear. In addition, the significance of the interactions among multiple process parameters on the weld forming quality is also obtained. Based on the results of the residual analysis, the linear regression model is modified in a targeted manner by adding the quadratic and interaction terms of the corresponding process parameters to establish a multivariate nonlinear regression model of the process parameters and the weld forming quality. Through comparing the goodness-of-fit, standard error, and standardized residual of the improved nonlinear regression model with the linear regression model and verifying the improved nonlinear regression model by verification experiment, it can be concluded that the improved nonlinear regression model has a better fitting precision and prediction precision. The results of the verification experiment indicate that the improved nonlinear regression model can correctly describe the relationship between the process parameters and the weld forming quality and can realize an accurate prediction of the weld forming quality.

Experimental Study on Physical and Mechanical Properties of Mixed Filling behind Retaining Wall

Through large-scale triaxial shear test and particle sieving test, the particle breakage characteristics and shear strength of mixed coarse-grained fillers behind a high filling wall in Guizhou Province were studied. The effects of mudstone content, particle gradation, relative density and stress state of coarse-grained mixed fillers on particle breakage and shear strength were analyzed. With the increase of mudstone content, gradation and relative density, the particle breakage rate decreases. With the increase of confining pressure and stress, the particle breakage rate of coarse-grained soil also increases. The shear strength cohesion increases first and then decreases with the increase of mudstone content, and the internal friction angle decreases gradually. With the improvement of particle size distribution and the increase of relative density, the cohesion and internal friction angle increase. At the same time, a multivariate non-linear regression model was established to synthesize the comprehensive internal friction angle. The predicted values were in agreement with the experimental data, and the optimum mixed packing parameters were selected.