Search

CN-122019926-A - Circumferential rate approximation method breaking through traditional numerical limit and high-end industrial calculation engine

CN122019926ACN 122019926 ACN122019926 ACN 122019926ACN-122019926-A

Abstract

The invention discloses a peripheral rate approximation method breaking through the traditional numerical limit and a high-end industrial computing engine, belonging to the technical fields of basic mathematical computation, high-end industrial computing power and precise industrial numerical computation. The method is based on unified theory of proportional philosophy origin and continuous interval, and is characterized in that GZ-PAT-011 is carried on to construct a specific continuous calculation interval of the circumference ratio, an adaptive super-convergence iteration formula is adopted to achieve ultra-high-bit circumference ratio approximation without error accumulation, a matched high-end industrial calculation engine adopts a nationwide industrial calculation power chip and a specific instruction set to convert a high-precision circumference ratio into an industrial precise calculation reference for ultra-precise industrial operation of aerospace, precise manufacturing, high-end equipment and the like. Compared with the traditional series method, the convergence speed of the invention is improved by more than 85%, the calculation of the unbiased circumference ratio to more than 1000 bits after decimal point can be realized, the industrial operation precision is improved by more than 90%, the bottleneck of the traditional numerical calculation is thoroughly broken through, and the autonomous and controllable high-end industrial calculation power is realized.

Inventors

  • Gui Wuyan

Assignees

  • 珠海市工正科技有限公司

Dates

Publication Date
20260512
Application Date
20260408

Claims (10)

  1. 1. A circumference ratio approximation method breaking through the limit of traditional numerical values is characterized by comprising the following steps of The golden section origin is calculated by defining a specific continuous calculation interval I_pi= [3.0, 3.5] of the peripheral rate in the proportion philosophy Center reference value C0, calculating by a Simpson integration method by adopting a continuous interval self-adaptive super-convergence iteration mode The interval integral value is used for gradually and iteratively updating the value of the circumference ratio, calculating the iteration error in real time and passing through a preset precision threshold value And (3) completing self calibration, triggering interval backtracking and re-iteration when the error exceeds the standard, and finally outputting a high-precision circumference ratio value.
  2. 2. The circumference ratio approaching method according to claim 1, wherein the origin center reference value C0 The calculation formula is C0=Lmin+phi× (Lmax-Lmin), wherein phi= (. V5-1)/2 is approximately equal to 0.618, and the iteration interval is equal to the sum of the values The width dynamically decays according to the rule wi=w (i-1) ×Φ.
  3. 3. The circumference ratio approximation method of claim 1, wherein the super-convergence iteration formula is Pi=pi (i-1) + (Wi/6) × [ f (ai) +4f (C0) +f (bi) ], wherein the function f (x) =4/(1+x2), Single point calculation errors are eliminated by integral operation.
  4. 4. The circumference ratio approximation method of claim 1, wherein the iterative error calculation formula is E= | (pi i-pi (i-1))/(pi (i-1) +10-24) |, mathematical level precision threshold is E < 10-18, high-end worker The industrial precision threshold value is E is less than or equal to 10 < -15 >.
  5. 5. The circumference ratio approximation method of claim 1, wherein the iteration termination condition is an interval The width converges to EPS stop=1e-18, or the error value reaches the corresponding accuracy threshold.
  6. 6. A high-end industrial computing engine for implementing the method of claim 1, comprising mathematics Local source interval modeling unit, super-convergence circumference ratio approximation unit, numerical adaptation conversion unit and domestic industrial special purpose The system belongs to a computing power chip, an industrial precise computing unit, an error real-time correction unit and an industrial data interaction unit.
  7. 7. The high-end industrial computing engine of claim 6, wherein the homemade industry-specific algorithm The force chip selects the rising 910B, loongson 3A6000 or domestic RISC-V incense core, and integrates the special instruction sets of PI_BOUNDS, PI_SIMPSON, PI_ATTEN, PI_CALIB and PI_ INDUST.
  8. 8. The high-end industrial computing engine of claim 6, wherein the numerical adaptation conversion unit The high-precision circumference ratio can be adapted to 15-bit, 30-bit and 50-bit different precision values according to industrial scenes, The adaptation formula is pi industrial =pi math ×s, S being the industrial scale factor.
  9. 9. The high-end industrial computing engine of claim 6, wherein the error real-time correction unit The industrial operational numerical deviation is automatically corrected by the hardware level logic according to the formula delta=k× Ecalib.
  10. 10. The high-end industrial computing engine of claim 6, wherein AI large can be directly embedded And the model is used for constructing a mathematical-level precision constant library, so that the reasoning precision of the scientific computing and industrial simulation AI model is improved.

Description

Circumferential rate approximation method breaking through traditional numerical limit and high-end industrial calculation engine The invention discloses a peripheral rate approximation method breaking through the limit of the traditional numerical value and a high-end industrial computing engine, belonging to the field of Basic mathematical computation, high-end industrial computing power and precise industrial numerical operation. The invention is based on proportional philosophy The unified theory of the source and the continuous interval is adopted, GZ-PAT-011 is adopted, the exclusive continuous calculation interval of the circumference ratio is constructed, An adaptive super-convergence iteration formula is adopted to realize the ultrahigh bit circumference rate approximation without error accumulation, and the method is matched with high-end industry The computing engine adopts nationally produced industrial computing power chip and exclusive instruction set to convert high-precision circumference value into industrial power The method is used for precision calculation standard and is used for ultra-precision industrial operation of aerospace, precision manufacturing, high-end equipment and the like. The invention The clear convergence rate is improved by more than 85 percent compared with the traditional series method, and the calculation of the unbiased circumference ratio to decimal point can be realized The industrial operation precision is improved by more than 90 percent after 1000 bits, the bottleneck of the traditional numerical calculation is thoroughly broken through, Realizing autonomous and controllable high-end industrial calculation power. Technical Field The basic mathematical value calculation, high-precision scientific calculation, high-end industrial power calculation engine and precise value The technical field of operation and AI large model reasoning optimization, in particular to a method based on proportional philosophy and continuous interval iteration The circumference ratio super approximation method of the above-mentioned method and a domestic high-end industrial computing engine for implementing the method. The invention is applicable to In aerospace, precision instruments, ultra-precision manufacturing, industrial simulation, measurement and control of high-end equipment and high AI large model Precision reasoning and the like have extremely demanding scenes of precision and stability. Background 1. The convergence speed of the traditional infinite series is slow, such as a Machin formula, a Chudnovsky formula, Gauss-Legendre iteration method, convergence speed is limited by discrete term number, and error is indicated in high-order calculation The number-level accumulation cannot break through the upper limit of the IEEE 754 double-precision value. 2. The existing algorithm does not introduce proportional philosophy origin logic, lacks continuous interval origin anchor points, and is easy to bias in the iterative process The distance source interval causes the calculation of the high-order circumference ratio to deviate, and the mathematical level accuracy requirement cannot be met. 3. Computing engine for high-end industrial computing, ultra-precise simulation and nanoscale measurement and control scene, and core dependence is overseas X86 ARM framework chip, domestic adaptation degree is not enough, and ultra-high precision industrial transportation with circumferential rate can not be supported Calculating the demand. 4. Industrial simulation, precision machining, radian calculation, fluid simulation and other scenes, lack of a mathematical origin calibration mechanism, The tiny numerical deviation of the circumference value can directly cause equipment manufacturing error, unstable simulation result and measurement and control precision Degree deviation problem. Disclosure of Invention The specific calculation system of the circumference ratio is constructed based on the proportion philosophy and the continuous interval, and the specific contents are as follows: 1. The source of the specific continuous interval of the peripheral rate is defined as I_pi= [3.0,3.5] which is the optimal stable initial interval under the industrial and mathematical dual application scene. Setting golden section The center reference value of the source is calculated by the formula of C0=Lmin+phi× (Lmax-Lmin), wherein Phi = (5-1)/2 ≡ 0.618, substituting into interval numerical calculation to obtain C0 ≡ 3.309017, which is used as circumference ratio And the core anchor points of iterative computation avoid the divergence problem in the iterative process. 2. The super-convergence circumferential rate approximates to an iteration rule, namely, a traditional discrete single point iteration mode is abandoned, and continuous interval self-adaption is adopted The iteration method is suitable for super-convergence, and the specific iteration rule is pi=pi (i-1) + (Wi/6) x [ f (ai) +4f (C0) +f (bi