Anvil cracking is a common accident during the synthesis of synthetic diamonds. Its causes can be categorized into normal failure and abnormal failure.
1、normal failure:
Material Characteristics: The anvil is primarily made of tungsten-cobalt cemented carbide with a cobalt content of less than 10%. This material, a skeletal polycrystalline alloy formed by sintering tungsten carbide as the matrix and cobalt as the binder, possesses high brittleness, high compressive strength, and wear resistance, but poor tensile and shear capabilities. Under alternating stress and thermal stress, tungsten carbide grains may slip, leading to loosening or fragmentation of the skeleton, or stress concentration at structural defects, thereby initiating microscopic cracks.
Structural Characteristics and Stress Conditions: When in operation, the anvil is subjected to alternating external load stress and alternating thermal stress. These stresses vary periodically, with a cycle frequency related to the diamond type and process time (ranging from tens of minutes to tens of hours), which is considered low-cycle fatigue. Even without considering local stress concentrations caused by defects, the anvil operates in an over-critical stress state, where fatigue damage and microscopic cracks will occur under alternating stress.
Research indicates the presence of tensile stress below the central axis of the anvil, and the maximum shear stress appearing at the upper one-third of the central axis. These maximum stress values have already exceeded the material's static strength limit and fatigue limit.
Large radial shear stress is distributed on the high-pressure anvil surface, which can lead to crack formation at the edges of the high-pressure anvil surface. These cracks continuously develop, propagating through many grains to become macroscopic cracks, ultimately leading to anvil cracking or crushing.
Visually, the anvil experiences transverse fracture and longitudinal cracking during operation.
Local Stress Concentration: The interference fit between the hammer head and the steel ring, along with a 1.5° inverse taper angle, causes uneven distribution of pre-tightening force on the outer cylindrical surface of the hammer head. This leads to local stress concentration, resulting in premature anvil damage. The maximum shear stress occurs on the central axis of the anvil, while the maximum tensile stress occurs on the symmetry line of the 46° inclined plane, causing premature failure to typically first occur on the 46° inclined surface of the anvil.
Temperature Field Influence: Alternating thermal stress is one of the main factors leading to anvil fatigue failure. Temperature changes cause material expansion or contraction, generating thermal stress when constrained. Furthermore, increasing temperatures also degrade the physical and mechanical properties of cemented carbide, such as compressive strength, hardness, and elastic modulus.
Internal Defects: Cemented carbide materials are extremely sensitive to defects. Once stress concentration occurs at a defect, no matter how small the stress, it will be the first point of failure.
Rapid Crack Propagation Speed: As a highly brittle material, once microscopic cracks appear in cemented carbide anvils, they rapidly extend and develop into macroscopic cracks, eventually leading to an "acoustic crack". This process is very fast.

2.Abnormal Failure:
Processing Technology and Quality Issues: Improper proportions of alloy elements like cobalt content, coarse grains, and other material defects, as well as delamination and cracks caused by processing technology, all reduce anvil performance, making them susceptible to damage.
Anvil Structural Defects: Unreasonable structural shapes, insufficient steel ring support, improper or inadequate hardness of the shims can lead to local stress concentration and anvil damage. Particularly, if the steel ring support is insufficient, it may cause rotation or detachment during operation, or even explosions under high temperature and pressure conditions, leading to anvil cracking.
Human Operational Factors: Improper anvil pressing and calibration can result in uneven stress distribution during pressure application. Poor alignment and synchronization of the press can lead to inconsistent depressurization, causing local stress concentration at the ends and edges of the anvil, damaging it and leading to "blasting".
Pyrophyllite Block Preparation Issues: Uneven preparation or excessive hardness of pyrophyllite blocks can reduce sealing performance. If the sealing pressure and the pressure inside the high-pressure cavity are not balanced, an explosion may occur, leading to "hammer extrusion" and anvil cracking.
Equipment Malfunction: Malfunctions of the cubic synthetic diamond press, such as water leakage causing rapid thermal expansion and contraction of the heating hammer, or failure of limit switches/displacement sensors leading to "hammer extrusion," can damage the anvil.
Improper Process: Setting heating power too high can lead to excessively high synthesis temperatures, or the presence of impurities at the electrical contact points can cause local thermal stress concentration and damage to the anvil.
Acoustic Emission-Based Online Detection Solution
Given the current lack of online detection products for anvil cracking both domestically and internationally, this research proposes an acoustic emission-based online detection solution for anvil cracking.
1.Rationale for Selecting Acoustic Emission Detection:
Principle: Acoustic emission (AE) technology detects AE signals generated by defects or abnormal structural regions under external or internal stress, to assess the severity of defects. When cemented carbide anvils are subjected to alternating loads, inherent material defects or micro-cracks formed during plastic deformation release strain energy as they propagate, generating AE signals. The formation and propagation of cracks produce strong AE signals, making them the primary AE source for detecting anvil cracking in cemented carbide anvils.
Advantages:
No Interference with Production: The detection process does not interfere with production. Detection components and front-end processing units can be encapsulated within a casing and mounted on the pre-tightening ring above the anvil.
Simple Detection Process: It only requires one detection transducer element and an additional front-end processing unit. Compared to other non-destructive testing methods, this significantly reduces equipment installation complexity and improves detection continuity and automation.
Low Cost of Signal Acquisition and Analysis: Existing press control system software and hardware resources (such as industrial PC sound cards) can be utilized for signal acquisition and processing, eliminating the need for expensive specialized equipment.
Energy Source: The energy originates from the intrinsic strain energy of the tested component, which better reflects internal characteristic changes of the signal, without requiring external energy.
Applicable Environment: It is insensitive to the geometric size and contact degree of the detection object, making it suitable for harsh environments such as high temperature, high pressure, and flammability.
2.Selection of Detection Signal Frequency Band:
Considering that experienced operators can identify anvil cracking by the "acoustic crack" phenomenon, and that macroscopic crack formation is accompanied by sound, the acoustic emission detection signal is determined to be in the audio band (20Hz ~ 20KHz). This allows for signal acquisition using an industrial PC sound card, reducing costs and equipment requirements. The detection goal is to promptly detect the expansion of microscopic cracks into macroscopic cracks or the formation of macroscopic cracks.
3.Framework of the Online Detection System: The system relies on the overall press control system, using its industrial PC as the core, and fully utilizing its software and hardware resources.
Sensor Detection Module:
Comprises a piezoelectric element, casing, damping agent, acoustic coupling agent, cables, and a pre-amplifier circuit.
YT-5 piezoelectric ceramic (diameter 14mm, thickness 0.8mm) is chosen for its high electromechanical coupling coefficient and sensitivity, suitable for audio band detection.
The sensor is installed using fastening bolts, secured in threaded holes on the upper part of the anvil to ensure tight contact.
The sensor casing is designed similarly to a pressure sensor, allowing the piezoelectric transducer element to be as close as possible to the anvil component. The interior is filled with an absorbent material and epoxy resin is used as an acoustic coupling agent to ensure good sound transmission and thermal insulation.
Signal Conditioning Module:
Performs pre-amplification, impedance matching, and filtering on the weak signal output by the sensor, allowing it to be directly input to the sound card.
Power supply design uses a LM1117-5V linear voltage regulator chip for internal power, with an external 7808 converting 24V down to 8V.
The pre-amplifier circuit uses a TLC2272AMD rail-to-rail dual operational amplifier to achieve high-impedance to low-impedance conversion and signal amplification, improving the signal-to-noise ratio.
The filtering circuit uses a Sallen-Key filter, with a passband designed around 85Hz~20KHz, used to eliminate low-frequency power interference and high-frequency electrical noise.
Data Acquisition Module:
Utilizes the industrial computer's sound card (e.g., ALC662 chip) line-in port for A/D conversion. Sound cards offer high precision, strong anti-interference capabilities, cost-effectiveness, and good compatibility. They use DMA for data transfer, reducing CPU usage.
The system software drives the sound card by calling Windows audio API functions, creating multiple audio buffers to enable continuous sampling and ensure uninterrupted data acquisition.
Signal Analysis Module:
Employs a C++ and Matlab hybrid programming approach. C++ handles sound card signal acquisition, graphical user interface display, and a multi-threaded signal analysis framework. Matlab is used to write signal recognition algorithms, which are then compiled into COM components for C++ to call.
A dual-threshold endpoint detection algorithm based on short-time energy and short-time average zero-crossing rate is designed to extract suspected signals and cracked hammer signals, performing signal pre-processing to improve recognition efficiency.
Feature analysis is conducted on the collected suspected signals and cracked hammer signals:
Time-domain analysis reveals similar energy and waveform contours between the two, with no significant differences.
Spectrum analysis (FFT) shows that suspected signals have distinct features in the low-frequency range below 5KHz (especially around 300Hz and 2KHz~4KHz). Cracked hammer signals, in addition to low-frequency components, also exhibit high-frequency components above 8KHz.
Time-frequency domain analysis (STFT) further demonstrates that suspected signal frequencies are generally below 5KHz, while cracked hammer signals, besides low-frequency components, show high-frequency components around 13KHz and 15KHz in the 12ms~15ms time segment.
EMD (Empirical Mode Decomposition) and HHT (Hilbert-Huang Transform) marginal spectrum analysis results are even clearer. Suspected signals exhibit distinct features in the 200Hz ~ 400Hz and 2KHz ~ 4KHz frequency bands. The fault frequency band for cracked hammer signals is found in the 13KHz ~ 16KHz range. EMD decomposition effectively reduces the influence of low-frequency components by subtracting local means, making fault frequency features more prominent.
Alarm Output Module:
Upon detecting an anvil crack signal, the alarm output module transmits the alarm signal to the press control system for timely fault operations such as stopping heating and depressurizing.
Currently, this detection system is still collecting signals on-site, and signal recognition algorithms are being improved based on existing sample features. Future goals include continued signal collection to increase sample diversity, in-depth theoretical research on signal analysis, testing in multiple industrial settings, and attempting to theoretically validate the reasonableness of the 13KHz ~ 16KHz characteristic frequency. Finally, considering the use of embedded platforms to improve system efficiency.