Abstract: Abstract: A biaxial rotary tillage machine is widely used in intelligent agriculture with ever-increasing demands in recent years in China, such as biaxial rotary tiller, biaxial stubble cutter, and biaxial ridging machine, because of its high operation efficiency, less soil compaction, and low energy consumption. A test bench is mostly utilized to simulate the field condition for the rotary tillage machinery, in order to reduce the rework times for higher design efficiency. Much effort has been made to explore the technology and theory of rotary tillage machines, especially commonly-used linear- and rotary-type indoor test benches. But there are still some limitations in these kinds of test bench at present, such as low universality, only suitable for a single axis rotary tillage test. Meanwhile, most test benches are fixed indoor installation covering a large area with low mobility and reliability. As such, the indoor test using remolded soil is difficult to reflect the real environment of operation sites. Therefore, it is highly demanding for a new test device to improve the performance of biaxial rotary tillage machines, according to the fundamental theory of single axis rotary tillage. In this study, a field mobile test bench was proposed to perform the throwing soil with the forward rotary blade roller, with the crushing soil with the backward crushing roller. These operations aimed to meet the high agronomic requirements of vegetable layered tillage, where the topsoil was fine soil, while the subsoil contained some large soil particles. A prototype of a biaxial ridge machine was then designed, together with the key components, to realize the real-time adjustment of relative space distance of two shafts, as well as the rotation speed ratio of front-rear roller shafts. A theoretical analysis was made on the working range of relative spatial position between two roller shafts, the structural parameters for the position adjustment device in the crushing roller shaft, and the motion parameters for the rotary speed adjustment device in the rotary blade shaft. A selection was completed on the hydraulic pump, motor, and valve, as well as on the torque sensor, PLC communication module, and data acquisition software. A detailed description was given on the working principle, test and data acquisition process of the whole machine. A central composite experimental design was adopted to improve the working quality of layered tillage, while reducing the energy consumption. A three-factor and five-level response surface experiment was conducted, where three influencing factors were taken as the horizontal and vertical distance between two shafts, as well as the shaft rotation speed of soil crushing roller, whereas, two indicators were set as the average power consumption, and the breakage ratio of soil at the top soil layer of 5 cm. Design-Expert software was then selected to analyze the test data. A regression model was established to clarify the significant correlation between each factor and indicator, aiming to comprehensively optimize the influencing factors. The results showed that three determinants were ranked in the order of horizontal distance between two shafts, vertical distance between two shafts, rotating speed of crushing roller shaft, in terms of impact on average power consumption, whereas, the determinants were ranked in the order of horizontal distance between two shafts, rotating speed of crushing roller shaft, vertical distance between two shafts, in terms of impact on the ratio of soil breakage. An optimal combination of work parameters was achieved as bellow: the average power consumption was 17.92 kW, and the ratio of soil breakage was 91.65%, particularly when the horizontal and vertical distances between two shafts were 570, and 96 mm, respectively, and the rotating speed of crushing roller shaft was 340 r/min. Specifically, the relative error between property indices was less than 5%, compared with the theoretical optimization and the valid test, indicating that the established model was expected to serve as the subsequent prediction and optimization. Meanwhile, the test data demonstrated that the developed test bench of biaxial rotary tillage was reasonable, while meeting the needs of multi factor and multi-level tests. The finding can provide a new potential test for the optimization design of similar components in the biaxial rotary tillage.